BRAIN‐online: An online cognitive and behavioral screening tool for fetal alcohol spectrum disorders

Abstract

Background

Fetal alcohol spectrum disorders (FASD) are among the most common developmental disorders but are underdiagnosed in part because of the inaccessibility of valid screening tools. The goal of this study was to test the ability of a novel online cognitive and behavioral screening tool to accurately distinguish youth with histories of prenatal alcohol exposure (PAE) from the comparison group.

Methods

Children were tested at home with an online cognitive and behavioral test battery, Brief Assessment of Individual Neurobehavior—online version (BRAIN‐online). Participants included children with (n = 239) and without (n = 89) histories of PAE between the ages of 5 and 17 years. BRAIN‐online includes a parent/caregiver‐completed behavioral questionnaire and seven cognitive tasks completed by the child. Performance data were analyzed using ANOVA, ROC analysis, and logistic regression.

Results

Children in the alcohol‐exposed group had more behavioral problems, as indicated by higher scores on the behavior rating, and performed more poorly on the cognitive tasks than children in the comparison group. Logistic regression results indicated that performance on BRAIN‐online distinguished the groups with a high degree of accuracy (accuracy = 80%, sensitivity = 79%, specificity = 83%, positive predictive value = 92%, negative predictive value = 60%). Results were not affected by attention‐deficit/hyperactivity disorder (ADHD) or overall intellectual ability.

Conclusion

BRAIN‐online is a novel online screening tool that can distinguish individuals affected by prenatal alcohol exposure from typically developing children with a high degree of accuracy. The availability of an accessible, low‐cost, efficient screening tool for fetal alcohol spectrum disorders (FASD) will improve diagnostic efficiency and reduce wait times and assessment costs.

Fetal alcohol spectrum disorders (FASD) are among the most common developmental disorders but are underdiagnosed because of the inaccessibility of valid screening tools. We describe BRAIN‐online, an accessible, web‐based, at‐home test that distinguishes individuals at risk for FASD with a high degree of accuracy (accuracy = 80%, sensitivity = 79%, specificity = 83%, positive predictive value = 92%, negative predictive value = 60%). The availability of an accessible, low‐cost, efficient screening tool for FASD will improve diagnostic efficiency and reduce wait times and assessment costs.

INTRODUCTION

Despite the prevalence of fetal alcohol spectrum disorders (FASD), challenges in detection have resulted in a high rate of misdiagnosis or missed diagnosis. The most recent study that documented a prevalence rate of 1%–5% in a large sample of first‐grade children reported that only 2 of the 222 children identified as having FASD by the study were previously known to have FASD (May et al., 2018). Another study in foster care indicated that 80.1% of children with FASD were not previously diagnosed, and 6.4% were misdiagnosed (Chasnoff et al., 2015). Diagnosis of FASD involves a multidisciplinary team, including a pediatrician, dysmorphologist or clinical geneticist, a psychologist or neuropsychologist, and other professionals (American Academy of Pediatrics, 2024). Limitations of the healthcare system and a lack of professionals trained in diagnostic procedures result in long wait times at a small number of specialty clinics (Popova et al., 2024). Affordable, accessible, and easy‐to‐use screening tools can enhance diagnostic efficiency by helping prioritize individuals who are most likely to benefit from comprehensive diagnostic evaluations, while potentially offering alternative recommendations for others.

Cognitive and behavioral impairment are hallmarks of FASD (Mattson et al., 2019), but assessment is time‐consuming, costly, and often inaccessible to people seeking diagnostic evaluations for FASD. In our previous studies (Goh et al., 2016; Mattson et al., 2023), we described a screening tool, the FASD‐Tree, with good sensitivity and specificity but that requires the involvement of a clinician skilled in physical measurements and does not include cognitive assessment. We aimed to develop a web‐based tool that could screen for cognitive and behavioral abilities known to be affected in individuals with FASD. In this paper, we describe the Brief Assessment of Individual Neurobehavior—online version (BRAIN‐online), which accomplishes this aim.

MATERIALS AND METHODS

Participants

Children aged 5–17 years were recruited from ongoing research studies at the Center for Behavioral Teratology at San Diego State University (SDSU), the Masonic Institute of the Developing Brain at the University of Minnesota, and through an online recruitment portal for FASD research from Indiana University. All three research centers were a part of the Collaborative Initiative on Fetal Alcohol Spectrum Disorders (CIFASD); data were collected between September 2020 and February 2025 (CIFASD4 and CIFASD5). A total of 357 children completed the study. We excluded participants because prenatal alcohol exposure (PAE) did not meet our inclusion criteria (see below; n = 15), invalid test data on two or more of the tasks (n = 2), or completion time >120 min (n = 12). The final sample comprised 328 participants.

Inclusion criteria were age (5–17 years), language (English speaking), and internet‐connected computer access. History of PAE was ascertained by maternal report, medical, legal, or social service records, and categorized as either (1) known or strongly suspected or (2) denied by parents or guardians. Participants with suspected PAE were included only if a reliable collateral report existed in the absence of direct maternal report or legal or social records. We compared participants with histories of PAE (the AE group, n = 239) to a typically developing comparison group (the CON group, n = 89). IQ testing was not a requirement for the study and participants recruited through the online portal were not asked to provide an IQ score. However, IQ testing was completed as part of the larger project for 181 participants (132 AE, 49 CON), and of these, 30 (22.7%) of the AE group had IQ scores <78 (i.e., 1.5 standard deviations [SD] below the population average). Similarly, 320 participants were asked about other diagnoses, and 112 (35%) reported current or previous diagnoses of attention‐deficit/hyperactivity disorder (ADHD).

Measure

BRAIN‐online is an online tool we developed to screen for the cognitive and behavioral effects of FASD. It includes a brief demographic survey, behavioral screen, and seven cognitive tests assessing fine motor function, reaction time, behavioral inhibition, attention, short‐term and long‐term spatial memory, spatial working memory, and executive functioning. Tests were gathered from open‐source material or developed to assess domains known to be affected in individuals with FASD. The BRAIN‐online Behavioral Questionnaire (BBQ) is completed by the parent and contains 14 behaviors chosen to discriminate between children with and without FASD. The demographic survey, BBQ, and seven cognitive tasks were administered using Gorilla Experiment Builder (www.gorilla.sc) (Anwyl‐Irvine et al., 2020). BRAIN‐online subtests are described in Table 1.

TABLE 1.

BRAIN‐online subtests. ^a

Task	Description	Measure included in analyses
BRAIN‐online Behavioral Questionnaire (BBQ)	Participant's parent or caregiver rates the participant on a list of 14 behaviors commonly observed in individuals with FASD (e.g., hyperactivity, acting young for age)	BBQ score (number of items endorsed)
Tapping	Participant taps the spacebar with their index finger as fast as possible for 10s. Trials are repeated five times for each hand	Average number of taps for dominant and nondominant hands
Reacting	Participant taps the spacebar in response to a centrally presented stimulus as quickly as possible	Median reaction time
Inhibiting	Participants respond to stimuli presented in a 2 × 2 grid. The task is to respond only when the stimulus is facing to the right and not when it is facing to the left	Accuracy (hits), false positive responses
Remembering	Participant is shown a 4 × 4 grid of 14 hidden playing cards and 2 blank spaces. In the exposure trial, each card is revealed consecutively. This is followed by three learning trials, an immediate recall trial, and a delayed recall trial after approximately 10–15 min. In all trials, participants are asked where they saw each of the cards, and responses are made by clicking the box on the grid corresponding to the location of the card. Feedback is provided on learning trials, and the correct location of the card is shown. In the recall trials, no feedback is given	Items recalled on the learning trials, immediate recall, delayed recall, retention (delay/immediate)
Fishing	Participant responds to a central stimulus that is presented with or without flanking stimuli. Flanking stimuli are either consistent or inconsistent with the central stimulus. The task is to indicate the direction of the central stimulus	Accuracy, reaction time for consistent, inconsistent, and neutral trials, flanker effect (difference between reaction time for inconsistent and consistent trials)
Stepping stones	Participant repeats a spatial span pattern that increases from 2 to 10 stimuli	Span length achieved, trials correct
Connecting	Participant connects circles in sequence in three trials: number order, letter order, or switching between numbers and letters in order	Time to completion for numbers, letters, switching trials

Note: BBQ, BRAIN‐online Behavioral Questionnaire; BRAIN‐online, Brief Assessment of Individual Neurobehavior.

^a Additional information about BRAIN‐online is available on request from the corresponding author.

Procedure

Parents of children enrolled in a multi‐site research program were emailed an invitation to participate in BRAIN‐online with a link to the online platform and a unique identifying code. Parents initiated BRAIN‐online, reviewed a consent document, and completed the demographic survey and BBQ. The child then completed the cognitive tests. Parents were allowed to help their child begin the test, including setting up the computer, reading the initial instructions, and prompting them to stay on task, but they were not to assist their child in completing the tests. Testing took an average of 44.9 min (SD = 12.90, range 20–102 min), and 90.5% of participants completed in less than 60 min. Participants were told they could take breaks between tasks but not after beginning the Remembering task to preserve the fixed time allocated for the delay period. Procedures were approved by the Institutional Review Board at SDSU.

Analyses

Data were analyzed using SPSS v. 29.0.2.0. For sample characteristics, unadjusted analyses were conducted using analysis of variance (ANOVA) for continuous data (e.g., age) and chi‐squared tests for categorical data (e.g., sex, race, and ethnicity). Prior to analysis, data were screened for outliers in SPSS and extreme data were omitted test‐wise. Classification accuracy was determined using chi‐squared tests, receiver operating characteristic (ROC) analysis, and sensitivity (true positive rate) and specificity (true negative rate) analyses. Indicators of test accuracy including positive predictive values (PPV) and negative predictive values (NPV) were calculated. Two‐sided tests with 95% confidence intervals (CIs) that exclude 1.0 or p‐values <0.05 were used to determine the statistical significance for all analyses.

RESULTS

Demographic information

Demographic data for both groups are presented in Table 2. The groups were similar in terms of age (F(1,326) = 0.178, p = 0.674), sex (X ² = 0.540, p = 0.463), gender identity (X ² = 3.911, p = 0.418), ethnicity (X ² = 1.543, p = 0.462), handedness (X ² = 4.963, p = 0.084), and family income (X ² = 0.88, p = 0.347). There were marginal but nonsignificant differences in race (X ² = 13.182, p = 0.068). Results from IQ testing were available for 181 participants and, as expected, the CON group had a higher mean IQ score than the AE group (F(1,180) = 54.700, p < 0.001). Low IQ scores (<78) occurred only in the AE group (22.7%).

TABLE 2.

Demographic characteristics for participants in the group with prenatal alcohol exposure (AE group) and the comparison group (CON group).

		CON (N = 89)		AE (N = 239)
		N	%	N	%
Sex	Female	48	53.9	118	49.4
	Male	41	46.1	121	50.6
Gender	Female	46	51.7	117	31.4
	Male	41	46.1	120	50.2
	Transgender/Does not identify	2	2.2	2	0.8
Race	American Indian/Alaska Native	0	0.0	11	4.6
	Asian	3	3.4	6	2.5
	Black or African American	3	3.4	30	12.6
	Native Hawaiian or Other Pacific Islander	0	0.0	1	0.4
	White	60	67.4	133	55.6
	More than one race	22	24.7	51	21.3
	Unknown/Not reported	1	1.1	7	3.0
Ethnicity	Hispanic or Latino	28	31.5	75	31.4
	Not Hispanic or Latino	59	66.3	151	63.2
	Unknown/Not reported	2	2.2	13	5.4
Handedness	Right	80	89.9	193	80.8
	Left	9	10.1	39	16.3
	Mixed	0	0.0	7	2.9
	Unknown/Not reported	0	0.0	0	0.0
IQ score a	IQ < 78	0	0.0	30	22.7
Family income b	<$50,000	7	17.5	22	25.0
ADHD c	Parent‐reported ADHD diagnosis	20	23.3	92	38.8

	CON		AE
	Mean	SD	Mean	SD
Age (years)	10.0	3.33	10.2	3.52
FSIQ (n = 182)	107.2	13.37	89.9	14.43

Abbreviation: ADHD, attention‐deficit/hyperactivity disorder.

^a IQ scores available for 182 (55.5%).

^b Income available for 128 (39.1%).

^c ADHD diagnoses available for 320 (97.6%).

Correlations with age and sex

Pearson correlations between age and BRAIN‐online performance were calculated. Age was significantly correlated (p < 0.05) with all but one subtest score and was included as a covariate in subsequent analyses. Correlations between sex and BRAIN‐online performance were calculated, and only one correlation was significant. To further explore potential sex differences, the main effect of sex, group, and the group x sex interaction were tested for all BRAIN‐online subtests. For nearly all analyses, the main effect and interactive effect of sex were not significant. Thus, sex was not included in subsequent analyses. Correlations between BRAIN‐online performance and age and sex are presented in Table 3. Correlations between all BRAIN‐online variables are included in a supplemental table (Table S1).

TABLE 3.

Pearson correlation coefficients between age or sex and performance on BRAIN‐online subtests.

Subtest	Measure	Age	Sex
BBQ	BBQ score	−0.124*	−0.022
Tapping	Dominant hand	0.550**	−0.002
	Nondominant hand	0.573**	−0.114*
Reacting	Median	−0.573**	0.095
Inhibiting	Hits	0.344**	−0.001
	False positives	0.220**	0.041
Remembering	Learning 1–3	0.391**	0.035
	Immediate	0.355**	0.025
	Delay	0.401**	0.101
	Retention	−0.026	0.052
Fishing	Accuracy	0.474**	0.065
	Flanker effect	−0.533**	−0.02
Stepping stones	Span length	0.502**	0.013
	Trials correct	0.549**	0.027
Connecting	Numbers	−0.463**	−0.025
	Letters	−0.534**	−0.04
	Switching	−0.598**	−0.096

Abbreviations: BBQ, BRAIN‐online Behavioral Questionnaire; BRAIN‐online, Brief Assessment of Individual Neurobehavior.

^* p < 0.05.

^** p < 0.001.

BRAIN‐online behavioral questionnaire (BBQ)

Descriptive data are presented in Table 4. For the BBQ, the number of items endorsed was summed for each participant and analyzed by ANOVA. The AE group had a higher average BBQ score than the CON group (F(1,325) = 153.573, p < 0.001). Age was a significant (p < 0.05) covariate. ROC analysis tested the cutoff score to best distinguish the groups. The results indicated excellent discrimination between the AE and CON groups with an area under the curve (AUC) of 0.859 (p < 0.001). Based on the ROC analyses, three cutoff scores were tested for classification accuracy (≥4, ≥5, ≥6). The classification results are presented in Table 5. Because BRAIN‐online is proposed as a screening tool, we selected a cutoff score of ≥4 for subsequent analyses based on its high sensitivity, PPV, and overall accuracy.

TABLE 4.

Performance on BRAIN‐online subtests for participants in the group with prenatal alcohol exposure (AE group) and the comparison group (CON group). p‐values represent the main effects of group and age as a covariate for each analysis.

Subtest	Measure	CON		AE		Group			Age
		Mn	SD	Mn	SD	F	df	p‐value	p‐value
BBQ	BBQ score	2.9	2.94	8.0	3.55	153.573	1, 325	<0.001	0.003
Tapping	Dominant hand	42.8	10.03	40.4	11.61	4.604	1, 292	0.033	<0.001
	Nondominant hand	37.8	9.82	35.9	10.47	4.423	1, 292	0.036	<0.001
Reacting	Median RT	301.8	135.65	340.8	151.19	8.310	1, 322	0.004	<0.001
Inhibiting	Hits	48.7	3.58	47.4	5.10	6.949	1, 315	0.009	<0.001
	False positives	43.9	5.71	43.7	6.73	0.117	1, 314	0.733	<0.001
Remembering	Learning 1–3	15.4	6.83	12.1	7.56	18.154	1, 322	<0.001	<0.001
	Immediate	6.3	3.02	4.7	3.37	18.541	1, 322	<0.001	<0.001
	Delay	5.5	2.85	4.2	3.11	16.805	1, 322	<0.001	<0.001
	Retention (%)	93.5	42.21	95.1	56.43	0.058	1, 306	0.809	0.654
Fishing	Accuracy (%)	84.4	16.15	67.8	26.55	42.664	1, 312	<0.001	<0.001
	Congruent RT	760.9	236.70	799.3	258.7	4.491	1, 312	0.035	<0.001
	Incongruent RT	797.5	271.74	850.3	297.94	4.393	1, 312	0.037	<0.001
	Neutral RT	721.1	220.23	759.2	246.00	2.483	1, 312	0.116	<0.001
	Flanker effect	32.8	71.59	850.3	297.94	0.201	1, 312	0.654	0.160
Stepping stones	Span length	4.9	1.78	4.2	1.68	17.117	1, 324	<0.001	<0.001
	Trials correct	6.7	2.82	5.4	2.78	21.362	1, 324	<0.001	<0.001
Connecting	Numbers	31,103.3	17,584.78	46,016.9	36,585.71	15.107	1, 306	<0.001	<0.001
	Letters	34,994.7	29,244.22	45,064.6	35,587.49	13.388	1, 306	<0.001	<0.001
	Switching	120,924.8	87,553.77	156,165.4	113,497.81	15.198	1, 306	<0.001	<0.001

Note: Bold font indicates significant p values.

Abbreviations: BBQ, BRAIN‐online Behavioral Questionnaire; BRAIN‐online, Brief Assessment of Individual Neurobehavior. Note: Bold font indicates significant p values.

TABLE 5.

Classification statistics for three cutoff scores on the BRAIN‐online Behavioral Questionnaire (BBQ) alone and overall performance on BRAIN‐online, including BBQ results.

Cut‐off score	BRAIN‐online behavioral questionnaire (%)			BRAIN‐online (%)
	≥4	≥5	≥6	With BBQ score	With BBQ ≥4
Sensitivity	86.6	79.9	73.6	78.8	80.8
Specificity	67.4	75.3	80.9	83.1	77.9
PPV	87.7	89.7	91.2	92.3	90.4
NPV	65.2	58.3	53.3	60.4	61.2
Overall accuracy	81.4	78.7	75.6	80.0	80.0

Abbreviations: BBQ, BRAIN‐online Behavioral Questionnaire; BRAIN‐online, Brief Assessment of Individual Neurobehavior; NPV, negative predictive value; PPV, positive predictive value.

Logistic regression tested the effect of BBQ ≥4 (yes vs. no) and age on the likelihood of being in the AE group. The model was statistically significant, X ² (2, N = 328) = 90.685, p < 0.001, explained 35% (Nagelkerke R ² = 0.350) of the variance in group membership, and correctly classified 81.4% of participants. Participants with BBQ scores of ≥4 were 14 times more likely to be in the AE group than the CON group (OR = 14.186, 95% CI = 7.85 to 25.64, p < 0.001). Age was not associated with group membership (OR = 1.06, 95% CI = 0.98 to 1.16, p = 0.166).

BRAIN‐online cognitive subtests

Data from the seven cognitive subtests were analyzed using ANOVA. Descriptive data for the cognitive subtests, including a summary of the ANOVA results, are presented in Table 4. The AE group performed significantly more poorly than the CON group on nearly all subtests. Group differences were not noted on false positive responses (Inhibiting), retention of learned material (Remembering), and flanker effect (Fishing).

Overall performance on BRAIN‐online

Following the univariate tests, logistic regression was used to test the effects of the combined performances of BRAIN‐online on the likelihood of being in the AE group. The following nine variables were included in the analysis: BBQ score, Tapping (nondominant hand), Reacting (median reaction time), Inhibiting (hits), Remembering (learning, delayed recall), Fishing (accuracy), Stepping Stones (maximum span), and Connecting (switching). Age was included as a covariate. The model was statistically significant, X ² (10, N = 275) = 133.089, p < 0.001, explained 55% (Nagelkerke R ² = 0.552) of the variance in group membership, and correctly classified 80% of participants. Age (p < 0.001) and BBQ score (p < 0.001) were significant predictors in the final model. This analysis was repeated using the cutoff score of ≥4 on the BBQ instead of the continuous BBQ score and showed similar results. Accuracy, Sensitivity, Specificity, PPV, and NPV are presented in Table 5.

Misclassified participants (i.e., CON participants classified as AE or AE participants classified as CON) were examined for systematic differences. In comparison with correctly classified participants, participants in the CON group that were misclassified as AE had lower IQ scores, higher BBQ scores, and lower accuracy on the Fishing subtest. In comparison with correctly classified participants, participants in the AE group that were misclassified as CON had higher IQ scores and lower BBQ scores, were faster on Reacting, more accurate on Inhibiting, recalled more items on the learning trials and delayed recall trial of Remembering, were more accurate on Fishing, had longer span lengths on Stepping Stones, and were faster on the switching trial of Connecting. Participants in the AE group who were misclassified as CON were more likely to be White than those who were correctly classified. No other demographic differences were found.

To investigate whether the BBQ score was driving results, the logistic regression was repeated with age and BBQ score entered on the first step and the remaining BRAIN‐online variables on the second step. The model including Age and BBQ score was significant [X ² (10, N = 273) = 112.857, p < 0.001], accounting for 49% of the variance in group membership and correctly classifying 75% of participants. The remaining BRAIN‐online variables accounted for a significant amount of additional variance (6%) [X ² (8, N = 273) = 18.429, p = 0.018], correctly classifying 80% of participants. Thus, while BBQ results accounted for the majority of variance in group differences in BRAIN‐online performance, cognitive performance added additional and significant variance to the model.

Exploratory analyses

Attention‐deficit/hyperactivity disorder (ADHD)

Given the high rates of ADHD in individuals with FASD (Fryer et al., 2007; Kambeitz et al., 2019; Peadon & Elliott, 2010), we explored the role of ADHD in the results presented above. Our comparison group included a small subgroup of participants with parent‐reported ADHD diagnoses (n = 20). We repeated the logistic regression, testing the likelihood of being in the AE group with the addition of a variable indicating parent‐reported ADHD. The model was statistically significant, X ² (11, N = 271) = 133.221, p < 0.001, explained 56% (Nagelkerke R ² = 0.561) of the variance in group membership, and correctly classified 83% of participants (sensitivity = 81.6%, specificity = 85.3%). Age (p < 0.001), BBQ score (p < 0.001), and total accuracy on Fishing (p = 0.034) were significant predictors in the final model. The ADHD diagnosis variable was not significant (p = 0.631). Average BBQ scores for participants with parent‐reported ADHD diagnoses were lower (Mn = 5.1, SD = 2.84) than those in the AE group overall (Mn = 8.0, SD = 3.55) and those in the AE group with parent‐reported ADHD diagnoses (n = 94; Mn = 8.5, SD = 3.18).

Overall cognitive ability

Overall ability, as measured by IQ, is also an important predictor in any test of cognitive abilities. We repeated the logistic regression on the likelihood of being in the AE group with the addition of IQ as a predictor. The model was statistically significant, X ² (11, N = 146) = 81.349, p < 0.001, explained 61% (Nagelkerke R ² = 0.611) of the variance in group membership, and correctly classified 84% of cases (sensitivity = 84.6%, specificity = 81.0%). BBQ score (p < 0.001) and IQ score (p = 0.006) were significant predictors in the final model. These results require confirmation in a larger sample as IQ scores were available in 55.5% of our sample.

FASD diagnosis

BRAIN‐online is not proposed as a diagnostic tool. However, it is informative to examine how performance on the screening tool differs between individuals with and without an FASD diagnosis. Diagnostic information about FASD was available for a subset of participants. Of these, 89 were controls and 142 were in the PAE group, and 123 of the PAE group met criteria for FASD. We repeated the logistic regression, including the 123 participants with FASD and 89 controls. The model was statistically significant, X ² (10, N = 177) = 121.823, p < 0.001, explained 67% (Nagelkerke R ² = 0.667) of the variance in group membership, and correctly classified 79.1% of participants (sensitivity = 71.0%, specificity = 89.6%). Age (p = 0.001), BBQ score (p < 0.001), and total accuracy on Fishing (p = 0.009) were significant predictors in the final model.

DISCUSSION

We tested a novel cognitive and behavioral screening tool designed to identify the need for further diagnostic evaluation of FASD. Performance on BRAIN‐online was highly accurate (80%) in distinguishing children with histories of PAE from the comparison group. Overall ability, as measured by IQ score, did contribute to results but did not explain the group difference on BRAIN‐online. Presence of an ADHD diagnosis did not explain the results. Performance on BRAIN‐online was also related to FASD diagnosis, which was available for a subset of the participants. The behavioral screen (the BBQ) was effective in distinguishing between the AE and CON groups. Using a cutoff score of 4 or more accurately identified 80% of participants, and individuals with a score of 4 or more were 14 times more likely to be in the AE group. BBQ scores accounted for a large amount of variance in group differences, and cognitive scores added a significant amount of variance resulting in the overall high classification accuracy. This finding is consistent with the inclusion of both behavioral and cognitive impairments in the diagnostic criteria for FASD.

In this study, positive predictive values and overall accuracy rates were very good, ranging from 88% to 92% and 76% to 81%, respectively. Negative predictive values were lower, ranging from 53% to 65%. However, it is not unexpected for NPV rates to be lower in screening tools. In fact, because BRAIN‐online is designed to screen for FASD and not to serve as a diagnostic tool, higher PPV and lower NPV rates are desirable. This balance between positive and negative predictive values allows some patients to be diverted from additional, comprehensive FASD diagnostic evaluations while reducing both false positive and false negative results.

Previous research has suggested that upwards of 80% of individuals with an FASD are not correctly diagnosed (Chasnoff et al., 2015, May et al., 2018). Many more individuals at risk for FASD are likely not evaluated. In the prevalence study conducted by May et al. (2018), only 2 of 222 individuals identified with an FASD had been previously identified. Underdiagnosis of FASD may be linked to challenges in identifying history of alcohol exposure and lack of awareness and/or education about FASD, resulting in an absence or delay of proper treatment and care (Benz et al., 2009). Improved screening tools, like BRAIN‐online, can reduce underdiagnosis of FASD by providing a means to screen for cognitive and behavioral difficulties associated with FASD. Misdiagnosis may rely more on the specificity of measures. Although BRAIN‐online is specific when the comparison group is typically developing youth, and our exploratory analyses were promising, more research is needed to test the specificity of the test when compared to other neurodevelopmental disorders.

BRAIN‐online is accessible and accurate and thus should be considered for use in FASD screening. There are multiple opportunities for screening in different settings where BRAIN‐online may prove valuable. And screening will differ in different settings, dependent on the expected prevalence of FASD or the goal of screening. In pediatric practices or school settings where most children do not have FASD, BRAIN‐online enables rapid, large‐scale screening, with those showing cognitive or behavioral difficulties referred for comprehensive evaluation. In FASD clinics or neuropsychological assessment clinics with long wait times, where the risk of FASD is higher, using BRAIN‐online may help prioritize patients for assessment or streamline the assessment process. Epidemiologic studies of large numbers of children to determine prevalence rates would benefit from a low‐cost, efficient mechanism for screening the cognitive and behavioral abilities known to be affected in individuals with FASD. Tests like BRAIN‐online might also be useful in clinical trials providing a low‐cost accessible measure to use at pre‐ and post‐treatment visits. Additional research is needed to measure the test–retest reliability and practice effects of repeated administration.

Importantly, history of PAE is a critical part of the FASD diagnostic process and confirmation of PAE is required for some but not all of the FASD diagnoses. BRAIN‐online is not a diagnostic tool and does not include a comprehensive evaluation of PAE. Rather BRAIN‐online was designed as a screening tool and includes only a single question about PAE. Individuals found to be at risk for FASD using BRAIN‐online should be referred for a comprehensive diagnostic evaluation that would include a detailed history of PAE.

We previously described a separate screening tool, the FASD‐Tree, which requires a physical examination and two parent questionnaires. Results of the FASD‐Tree have good sensitivity and specificity in identifying individuals with FASD among nonexposed comparison participants (Mattson et al., 2023). While the accuracy rates of the FASD‐Tree are similar to those of BRAIN‐online, the FASD‐Tree requires a physical examination, which limits its accessibility. In addition, although FASD‐Tree results are related to neuropsychological outcomes (Hyland et al., 2023), the tool does not include screening for cognitive ability, which is an important part of the FASD diagnosis. The combination of FASD‐Tree with BRAIN‐online should be investigated. Preliminary analysis of a subgroup of participants with both measures indicates a combined accuracy of over 95%. This finding requires confirmation in a larger sample.

Our analysis suggested that having ADHD did not account for the group differences in BRAIN‐online performance. However, our comparison group included only a small number of children with ADHD, and additional research with a larger group of children with ADHD or other neurodevelopmental disorders, with and without PAE, is warranted. Future research should also investigate how well BRAIN‐online performance maps onto neuropsychological performance when tested in the traditional, in‐person manner. Preliminary results of ongoing analyses indicate that performance on BRAIN‐online subtests is correlated with performance on parallel in‐person neuropsychological assessment measures. Furthermore, the majority of BRAIN‐online subtests are equivalent to these neuropsychological measures, suggesting that both BRAIN‐online and in‐person measures are similar in their measure of certain cognitive domains.

In‐person neuropsychological testing remains the standard for assessing cognitive and behavioral performance; however, due to improvements in accessibility, affordability, and efficiency, computerized testing has gained popularity (Bauer et al., 2012). Computerized screening tools have effectively assessed cognitive and behavioral skills in both clinical and nonclinical groups, but few tools are designed to be completed remotely (Dwolatzky et al., 2004; Talebi et al., 2020). BRAIN‐online has the added benefit of at‐home assessment, which provides a means to streamline the expensive, time‐consuming neuropsychological assessment. We do not suggest that BRAIN‐online should replace in‐person assessment as part of the FASD diagnostic process. However, wait times for diagnostic evaluations are as long as 54 months (Popova et al., 2024), assessment is expensive (McDowell Group, 2020), and specialty clinics are rare. Cost‐effective and readily accessible online screening tools, like BRAIN‐online, may help reduce these barriers. Poor performance on BRAIN‐online may indicate that a comprehensive, multidisciplinary diagnostic evaluation should be prioritized. In contrast, strong performance on BRAIN‐online, especially if combined with behavioral questionnaire responses in the typical/nonclinical range, might suggest that a full FASD diagnostic assessment may not be needed. Additional evaluation of the criterion validity of BRAIN‐online is underway, with results expected to provide important information about the relationship between performance on BRAIN‐online and traditional neuropsychological tests.

While BRAIN‐online has many advantages, important limitations should be considered. Our results indicated that the behavioral questionnaire accounted for a large amount of explained variance. The cognitive subtests added a relatively small amount of explained variance, although this change was significant. It may be that the cognitive subtests have limited added value; however, there are multiple reasons to retain both behavioral and cognitive assessment. First, both cognitive and behavioral impairments are part of the clinical picture of FASD; thus, assessing them has ecological validity and provides useful information about the individual. In addition, for children, the behavioral portion of BRAIN‐online is parent‐completed while the cognitive portion is completed by the individual. Although parent report is a routine part of child evaluation, parents have knowledge of the child's PAE, and their responses may reflect this knowledge. The combination of both child‐ and parent‐completed measures allows for both direct and indirect measures of ability, strengthening the validity of the screening tool. Finally, future studies may reveal that performance on the cognitive measures provides important discriminatory power when comparing individuals with FASD to those with other neurodevelopmental disorders.

Accessibility of BRAIN‐online is currently limited to those who speak English and have an internet‐connected computer or tablet with an attached keyboard. Language of administration has a significant impact on performance among Spanish‐speaking bilingual individuals (Gasquoine et al., 2007). Linguistic adaptations of these measures are needed to strengthen data accuracy and interpretation of these results among Spanish‐speakers. We recently translated BRAIN‐online into Spanish, and future studies will describe BRAIN‐online performance of Spanish‐speaking adults and children. This study also focused on children aged 5–17 years and did not include adults. FASD in adults is understudied. Importantly, BRAIN‐online was developed for use across the age range, and future studies will describe performance in adults.

In summary, we describe herein results from a web‐based screening tool for FASD. Results indicate that performance on BRAIN‐online can be used to distinguish individuals affected by prenatal alcohol exposure from typically developing children with a high degree of accuracy. The availability of an accessible, low‐cost, efficient screening tool for FASD, such as BRAIN‐online, will improve diagnostic efficiency and reduce wait times and assessment costs.

CONFLICT OF INTEREST STATEMENT

The authors have no financial or other conflicts of interest.

Supporting information

Table S1

DATA AVAILABILITY STATEMENT

Additional information about CIFASD, including about data sharing opportunities, can be found at www.cifasd.org. Data from this study are part of Phase 4 (2017‐2022) and Phase 5 (2022–2027).