Abstract
Objectives
Early detection of cognitive decline is critical for timely intervention, yet traditional neuropsychological assessments are time-intensive, require trained examiners, and may introduce administration bias. Digital platforms can offer scalable and accessible alternatives by enabling automated scoring and self-administration. To address this gap, we developed Cognosis, a tablet-based cognitive battery tailored for older adults, and aimed to (1) provide preliminary demographically stratified reference data and (2) examine the effects of age, education, and gender on cognitive performance in cognitively healthy Korean older adults.
Methods
We conducted a community-based normative study using a cross-sectional design. A total of 381 cognitively healthy adults aged 55 to 84 years were recruited from community centers, hospitals, and other local settings across Seoul, Gyeonggi, and Jeolla provinces in South Korea. Cognosis assesses four cognitive domains: attention/executive function, language, memory, and social cognition. To evaluate demographic influences, we applied penalized regression models (Least Absolute Shrinkage and Selection Operator) and stratified the preliminary normative data by age (55–64, 65–74, and 75–84), education (0–6, 7–12, 13–16, and ≥17 years), and gender.
Results
Age and education significantly influenced cognitive performance, with younger age and higher education associated with better scores. However, memory performance was not associated with these demographic variables. While gender did not have a main or interaction effect in regression analysis, subgroup analyses revealed that highly educated women outperformed men in several cognitive domains, including attention/executive function, fluency, and verbal/visual memory.
Conclusions
This study provides preliminary, demographically stratified reference data for Cognosis, improving its clinical utility for cognitive assessment in Korean older adults. Future research should validate its applicability in clinical and cross-cultural populations.
Introduction
Dementia poses a substantial public health challenge, affecting nearly 55 million individuals worldwide. 1 The necessity is particularly pronounced in South Korea, where the rapidly aging population has led to an increased prevalence of neurodegenerative diseases, including dementia, which affects approximately 1 in 10 elderly individuals. 2 Despite significant advances in clinical research, accurate and early diagnosis of dementia remains challenging, especially in cases of mild cognitive impairment, where symptoms are often subtle and variable. 3 Furthermore, dementia subtypes such as Alzheimer's disease, vascular dementia, frontotemporal dementia, and dementia with Lewy bodies frequently present overlapping and coexisting symptoms, contributing to considerable diagnostic ambiguity.4,5
In current clinical settings, traditional cognitive assessment tools, including the Mini-Mental State Examination (MMSE), 6 the Korean version of the Consortium to Establish a Registry for Alzheimer's Disease (CERAD-K), 7 and the Seoul Neuropsychological Screening Battery-II (SNSB-II) 8 —are widely used. However, practical challenges arise; these traditional assessments require trained expertise, are time-consuming, and increase the workload of medical staff. Furthermore, conventional methods carry the risk of administration bias, whereas digital assessments demonstrate higher diagnostic accuracy and may reduce this risk. 9
Indeed, a recent review indicated growing interest in developing digital assessments to improve scalability, standardization, and feasibility in real-world settings. 10 However, much of the development and validation to date has been concentrated in Western, English-speaking settings, and cross-cultural generalizability is still limited. Accordingly, digital batteries designed for cross-cultural use have been proposed and evaluated in linguistically and culturally diverse contexts (e.g., Oxford Cognitive Testing Portal; OCTAL). 11 However, there still remains a need for digital assessments that are linguistically and culturally tailored to older adults and accompanied by population-specific reference data to support clinical interpretation.
To address this gap, we developed Cognosis, a digital neuropsychological assessment platform tailored for older adults in South Korea. Cognosis offers voice-guided instructions and an intuitive interface that enables unsupervised self-administration. The system facilitates automated scoring, and it aims to achieve a high degree of sensitivity in differentiating between diverse dementia subtypes by assessing various types of functions, including attention, executive function, memory, language, and social cognition. Its comprehensive integration of the workflow, spanning from administration to scoring, within a unified digital environment, is designed to alleviate clinician burden and enhance consistency.
In this study, we derived preliminary reference data from a community-based sample of Korean older adults to support the clinical validity of this tool. The effects of demographic variables—including age, education, gender, and their interactions—were modeled to enhance interpretability and fairness across diverse populations.12,13
Methods
Participants
The study collected data from 471 healthy elderly participants aged 55 to 84 between December 2022 and April 2024. To minimize selection bias, we recruited participants across a range of ages and educational levels and from both urban and rural settings. Recruitment was conducted at multiple community-based and institutional settings across Seoul, Chungcheong, and Jeolla provinces in South Korea. The full list of data-collection sites is provided in Supplemental Table S1. Exclusion criteria included any self- or informant-reported cognitive decline, lower MMSE for Koreans (MMSE-KC) scores than z-score −1.0 standard deviations, and a diagnosis of mild cognitive impairment or dementia. Additionally, individuals meeting the health exclusion criteria outlined by Christensen et al. 14 were excluded. Finally, we eliminated participants with response patterns suggesting a lack of understanding or concentration in at least two tests (e.g. providing the same “yes” or “no” response in more than 25 out of 30 questions in the visual/verbal memory test or generating fewer than 10 responses in the digit-symbol matching test). From the initial 472 participants, 21 participants were removed due to low MMSE-KC scores, 33 participants for not meeting the age eligibility criteria, and 13 participants for failing the response validity criteria, resulting in a final sample size of 404. Written informed consent was obtained from each subject before study participation. The Institutional Review Board of Cha University approved the study protocol (IRB No. 1044308-202209-HR-053).
Development of Cognosis
Cognosis has been developed to empower older adults to conduct assessments independently, facilitated by voice-guided instructions, screen layouts optimized for elderly users, and tutorials employing finger-image demonstrations alongside structured practice sessions. The interface has been designed to enhance usability by placing instructions at the top, test stimuli in the center, and intuitive response buttons at the bottom (Figure 1). Furthermore, the incorporation of distinct screen designs serves to distinctly differentiate the instruction and practice phases from the primary assessment phase, thereby enhancing the clarity and intuitiveness of the user experience. The test instructions and items are presented sequentially and automatically, requiring no manual navigation by the user. It also includes adaptive feedback during certain tasks: if a participant makes three or more consecutive errors, the correct answer is shown before continuing. However, this skip policy does not apply in situations where user input is essential for scoring.
Example screens from Cognosis (Digit Symbol Substitution Test).
The test is administered using an application installed on a tablet. It is designed for compatibility with tablet devices meeting the recommended specifications—iPad (iPad 8 or later, running iOS 15.5 or higher), and requires a Wi-Fi or cellular connection for optimal performance. Response data are temporarily stored on the participant's local device throughout the session and are securely transmitted to the central server upon completion of the final item. After transmission, the results are automatically scored and made available to authorized clinicians/study staff via an internal results interface. Results were reviewed and communicated to participants by the clinician as appropriate, and a copy of the score output was provided upon request.
It shows the assessment phase of the Digit Symbol Substitution Test, where stimuli and response buttons are clearly arranged (instruction at the top, stimuli in the center, and responses at the bottom) to maximize usability and independence.
Assessments of Cognosis
Cognosis is composed of 10 subtests, which are divided into 5 distinct domains. These subtests are designed to assist in the differentiation of dementia subtypes other than AD. The total time required to complete the assessment is approximately 1 hour. A detailed description and figures of each test can be found in the Supplemental material (Figures S1–S9).
Attention/executive function
The Stroop Test15,16 measures the ability to maintain focus while inhibiting irrelevant responses. The Trail Making Test 17 assesses a person's ability to connect elements in a given order, where Part A connects only numbers, and Part B requires connecting numbers and days of the week. The Digit Symbol Substitution Test 18 assesses processing speed and accuracy in matching numbers to corresponding symbols.
Language and semantic memory
The Fluency Test 19 measures a participant's ability to generate words in the categories “animals” and “fruits” within 30 seconds per category. The Size/Weight Attribute Test 20 assesses a participant's ability to rank stimuli by size and weight, where participants arrange animals from largest to smallest and objects from heaviest to lightest.
Verbal/visual memory
The Verbal Cued Memory Test 21 assesses recall and recognition of learned verbal stimuli. Participants indicate whether each word was previously learned by selecting “Yes” or “No”. Visual Pattern Memory Test 22 assesses the ability to recognize and differentiate between previously viewed and new visual stimuli using nonverbal pictured items. Participants indicate whether each picture was previously learned by selecting “Yes” or “No.”
Social cognition
The Theory of Mind Test 23 evaluates a participant's ability to understand emotions and social contexts within narratives. The Facial Emotion Recognition Test 24 assesses a participant's ability to identify emotions from facial expressions. Participants selected one of six emotion categories: “Joy,” “Surprise,” “Sadness,” “Fear,” “Anger,” and “Neutral.”
Other clinical assessments
Several standardized clinical assessments were used in this study. To screen for cognitive decline, we administered the CERAD-K, 7 including the MMSE-KC. 25 Both CERAD-K and MMSE-KC assess various cognitive domains, with lower scores indicating greater impairment. Depressive symptoms were measured using the Korean Version of the Short Form Geriatric Depression Scale (SGDS-K), 26 a 15-item yes/no questionnaire, where higher scores indicate more severe depressive symptoms.
Statistical analysis
To provide preliminary evidence of convergent validity of Cognosis, Pearson correlations were computed between Cognosis scores and corresponding MMSE and CERAD-K indices. p-values were adjusted using the Benjamini–Hochberg procedure to control the False Discovery Rate (FDR).
To assess the relative contribution of age, education, and gender to each test score, we employed the Least Absolute Shrinkage and Selection Operator (LASSO). 27 We identified and handled outliers using the interquartile range (IQR) method, where values beyond 1.5 × IQR from the first and third quartiles were replaced with missing values. Additionally, responses that fell below the expected range for cognitively healthy older adults, as determined by a psychiatrist and a clinical psychologist, were considered outliers and removed (i.e. scoring below 6 on the visual/verbal memory tests or Digit Symbol Substitution Test). The penalized regression model is widely used for its efficiency and sparsity in variable selection. In the analyses, each test score served as the dependent variable, while age, education level, gender, and their interaction terms were included as predictor variables. Age and education were treated as continuous variables, and gender was coded as 0 for women and 1 for men. LASSO applies an L1 penalty, which constrains the sum of the absolute values of the regression coefficients, thereby performing both variable selection and regularization. This method automatically selects the most predictive variables while shrinking the coefficients of less relevant variables toward zero, effectively mitigating issues related to multicollinearity (variance inflation factor). Given that interaction terms can exacerbate multicollinearity, LASSO was particularly well-suited for the present analysis. We used repeated random split validation to improve robustness to random data partitioning. Specifically, for each outcome, the analytic dataset was randomly split into a training set and a test set (train–test split = 0.3), and this procedure was repeated 500 times with different random splits. Within each training set, the regularization parameter (λ) was selected via 10-fold cross-validation, using the value within one standard error of the minimum cross-validated error (1-SE λ). The fitted model was then evaluated on the corresponding test set. Coefficient estimates were summarized across the 500 repetitions by reporting the mean coefficient and the empirical 2.5th and 97.5th percentiles. Predictors whose 95% interval excluded zero were considered to have survived across splits, whereas predictors with intervals including zero were treated as nonsurviving. All models were implemented using the glmnet package in R, with predictors standardized internally for penalization (standardize = TRUE); coefficients are reported on the original scale of each predictor.
Based on this analysis, we derived age- and education-stratified reference values. To examine gender differences within each age and education group, we conducted two-sample t-tests (two-tailed) with a significance level of 0.05. All statistical analyses were performed in R (version 4.4.0). 28
Results
Demographic characteristics
The demographic characteristics of the subjects are presented in Table 1.
Demographic table.
CERAD-K: Korean version of the Consortium to Establish a Registry for Alzheimer's Disease; MMSE-KC: Mini-Mental State Examination in the Korean version of CERAD Assessment Packet; SGDS: Short Form Geriatric Depression Scale.
Correlations with MMSE and CERAD-K
Correlations between Cognosis and conventional cognitive measures are presented in Supplemental Table S2 and Figure S10. MMSE-KC and CERAD-K total scores showed significant positive associations with all subscores of Cognosis after FDR correction. The largest correlations were observed for the Stroop Test (word: r = 0.541, p < 0.001, p FDR < 0.001; word-color: r = 0.575, p < 0.001, p FDR < 0.001), Digit Symbol Substitution Test (r = 0.533, p < 0.001, p FDR < 0.001), and Fluency test (r = 0.578, p < 0.001, p FDR < 0.001). Correlations were also consistently positive between conceptually aligned subtests (Supplemental Table S2).
Effects of age, education, and gender on the test score
The LASSO analysis identified age, education, and gender as predictors of test performance across multiple cognitive domains (Table 2). In the attention/executive function domain, younger age was associated with better performance on the Stroop Test (word: β = −0.135, 95% confidence interval (CI) [−0.163, −0.107]; word-color: β = −0.161, 95% CI [−0.209, −0.116]), Trail Making Test (trails A: β = 0.108, 95% CI [0.074, 0.139]; trails B: β = 0.147, 95% CI [0.045, 0.256]), and Digit Symbol Substitution Test (β = −0.174, 95% CI [−0.211, −0.137]). Higher education was also linked to better performance on the Stroop Test (word: β = 0.199, 95% CI [0.148, 0.252]; word-color: β = 0.337, 95% CI [0.261, 0.423]), Trail Making Test B (trails A: β = −0.176, 95% CI [−0.240, −0.107]; trails B: β = −0.408, 95% CI [−0.584, −0.232]), and Digit Symbol Substitution Test (β = 0.233, 95% CI [0.165, 0.299]).
Selected features and coefficients from LASSO regression models.
The 95% confidence intervals (CI) were derived from 500 repeated train–test splits.
In the language and semantic memory domain, education was positively associated with performance on the Fluency Test (β = 0.117, 95% CI [0.027, 0.204]), while age was negatively associated (β = −0.098, 95% CI [−0.148, −0.047]). In the memory domain, both verbal and visual memory were not associated with age, gender, education, or interactions. Finally, in the social cognition domain, age was associated with poorer performance on the Facial Emotion Test (β = −0.061, 95% CI [−0.092, −0.027]), and education was associated with better performance on the Theory of Mind (β = 0.033, 95% CI [0.006, 0.060]).
Preliminary normative data
Based on the analysis of demographic variables, reference values for test scores were summarized by age (i.e. 55–64, 65–74, and 75–84 years of age), years of education (i.e. 0–6, 7–12, 13–16, and longer than 17 years), and gender (Table 3). The means, standard deviations, medians, 5th percentile, and 95th percentile for each subtest are presented in Supplemental material.
Sample sizes by demographic strata (age × years of education × gender).
Regarding assessments of attention and executive function, the mean number of correct responses on both the word and word-color trials of the Stroop Test increased as participants had more years of education and were younger (Supplemental Tables S3–S4). For the word trial, female participants demonstrated significantly higher accuracy compared to male participants in the 75 to 84 age group with over 17 years of education (t = −3.09, p = 0.01). On the word-color trial, female participants aged 55 to 64 showed significantly higher accuracy than their male counterparts with 0 to 6 years of education (t = −3.41, p = 0.02). Similarly, females aged 75 to 84 exhibited significantly higher accuracy than males in the 7 to 12 (t = −2.36, p = 0.03) and over 17 years of education groups (t = −2.85, p = 0.02). Overall, the response time for Trail Making Tests A and B increased as participants had more years of education (Supplemental Tables S5–S6). For trails A, scores improved as age decreased, though no distinct pattern emerged for trails B. For trails A, a significant gender difference emerged among those with over 17 years of education, where male participants took longer to complete the task compared to females aged 55 to 64 (t = 2.92, p = 0.007) and 65 to 74 (t = 2.43, p = 0.02). The same pattern was also found in those aged 75 to 84 with 7 to 12 years of education (t = 2.06, p = 0.05). On trails B, females aged 75 to 84 with 7 to 12 years of education completed the task significantly faster than males (t = 2.24, p = 0.04). The mean accuracy on the Digit Symbol Substitution Test increased as participants had more years of education and were younger, with no significant gender differences observed across subsets (Supplemental Table S7).
For tests evaluating language abilities, the mean accuracy on the Fluency Test increased with longer years of education and younger age (Supplemental Table S8). In the 55 to 64 age group, significant differences emerged between male and female participants with 0 to 6 years (t = −3.01, p = 0.04) and over 17 years of education (t = −3.01, p = 0.006). Regarding the Size and Weight Test, no clear pattern was observed in scores as a function of age or education years (Supplemental Table S9). Male participants demonstrated significantly superior performance compared to females in the 65 to 74 age group with 7 to 12 years of education (t = 2.40, p = 0.02). A similar trend of better performance by male participants was also found in the 75 to 84 age group with 0 to 6 years of education (t = 3.60, p = 0.001).
Regarding memory function, the mean Visual Cued Memory Test scores increased with more years of education in the youngest age group (55–64), but this pattern was not observed in the older age groups (65–74 and 75–84) (Supplemental Table S10). Furthermore, the mean Verbal Cued Memory Test scores did not exhibit any clear trends related to age or education level (Supplemental Table S11). However, female participants between the ages of 55 and 64 with more than 17 years of education (t = −2.63, p = 0.01), as well as those aged 65 to 74 with 13 to 16 years of education (t = −3.45, p = 0.003), performed significantly better than their male peers on this measure.
Lastly, in the social cognition domain, performance on the Theory of Mind Test was negatively associated with age, with younger participants demonstrating higher accuracy (Supplemental Table S12). Among individuals aged 65 and older, accuracy increased with years of education, though a slight decline was observed in those with more than 17 years of education. In terms of gender differences, male participants aged 75 to 84 with 0 to 6 years of education outperformed their female counterparts (t = 2.13, p = 0.04). Similarly, in the Facial Emotion Test, performance generally declined with age and improved with higher education levels (Supplemental Table S13). Additionally, female participants outperformed males in certain subgroups, specifically in the 55 to 64 age group with more than 17 years of education (t = −2.17, p = 0.04), the 75 to 84 age group with 13 to 16 years of education (t = −3.15, p = 0.004), and the overall 75 to 84 age group (t = −3.72, p = 0.002).
Discussion
This study aimed to provide preliminary, demographically stratified reference data for the newly developed digital cognitive assessment battery, Cognosis, and to examine associations of age, gender, and education with Cognosis test performance. The results demonstrated that age and education were consistently associated with test performance across multiple cognitive domains, including attention/executive function, language, and social cognition. To support the interpretation of Cognosis in the Korean elderly population, we stratified the preliminary normative data based on these demographic variables.
Through an automated feature selection process, we found that age was positively associated, while years of education were negatively associated with test performance in the domains of attention/executive function, language, and social cognition. These findings align with existing literature demonstrating the significant influence of age and education on cognitive test performance, where older age is typically linked to lower cognitive abilities, while higher educational attainment is associated with better performance.13,29–32 To be specific, a systematic review found that cognitive decline was common in aging, with evidence from multiple longitudinal studies showing that the prevalence and rate of cognitive impairment increase with age. 29 Also, based on a large-scale review suggested that cognitive abilities generally decline with age, though the rate of decline varies across domains. 30 For education, it was identified as one of the most significant predictors of late-life cognition in a large sample of nearly 1000 older adults. 31 This finding is further supported by a comprehensive review incorporating evidence from longitudinal cohort studies and meta-analyses, highlighting the strong association between higher educational attainment and better cognitive function in old age. 32 This underscores the necessity of implementing demographic adjustments to the normative data for this newly developed neuropsychological assessment battery.
In contrast, performance on the verbal and visual memory test did not show a significant association with demographic factors or their interactions. This finding diverged from previous research indicating that younger age and higher educational attainment are linked to enhanced episodic memory, working memory, and semantic memory abilities, particularly among older adults.29,31,32 This discrepancy may be attributed to the exclusion of participants with cognitive impairment, as identified through measures used to diagnose dementia (e.g. MMSE-KC and CERAD-K), which is typically characterized by pronounced memory deficits. Additionally, the difference in assessment protocols between the current and traditional methods may have contributed to this outcome. These findings highlight the need for further research to clarify the distinct cognitive mechanisms underlying visual and verbal memory performance and to explore their differential roles as predictors of cognitive function.
This study revealed a complex and context-dependent relationship between gender and cognitive performance. While prior research has shown that men outperform women in perceptual reasoning, working memory, and spatial memory,33–35 whereas women excel in processing speed, emotion recognition, and verbal memory,36–38 the current findings were more nuanced. Penalized regression analysis did not retain gender as a non-zero main or interaction term, suggesting that any gender-related differences were not consistently expressed as linear effects across the full sample. However, when scores were examined within age- and education-stratified subgroups as part of the normative analyses, several comparisons showed differences between men and women. Specifically, female participants outperformed their male counterparts on various cognitive tests, including those assessing attention/executive function, language fluency, and verbal/visual memory. This finding was partially consistent with previous literature, as women demonstrated superior performance in verbal memory and fluency tasks, as expected. However, the study also found that men did not outperform women in visual memory. We observed that female participants with higher levels of education tended to exhibit superior cognitive performance in many of those tests. This was particularly relevant given the cultural context in South Korea, where historically, female elderly individuals had fewer opportunities to attain formal education compared to their male peers. 39 Therefore, the superior performance of highly educated female participants may reflect the influence of other social and individual factors, such as higher socioeconomic status 39 or intellectual capabilities. Further investigation is warranted to track the persistence of these gender-based disparities as societal norms and gender equality in South Korea continue to evolve.
Limitations
This study has several limitations. First, the unsupervised nature of the digital assessment may have introduced variability in data quality and made it more difficult to verify participant compliance. To enhance the reliability of future assessments, incorporating validity indicators—such as dummy items or attention checks—could help identify potential inconsistencies while preserving the benefits of the digital format. Second, although participants were recruited from multiple regions to capture demographic and geographical variability, the sample was not drawn using probability-based methods and therefore may not reflect national population distributions, including socioeconomic composition. Also, as the study was conducted in South Korea, the generalizability of the findings may be limited to older Korean adults. As such, further research is warranted to compare sample characteristics with national census distributions and evaluate the extent to which the findings generalize across different linguistic and cultural contexts. Furthermore, some age × education × gender strata had small cell counts; therefore, the stratified reference values should be interpreted as preliminary. Larger samples with adequate numbers per stratum (e.g. 50–75 individuals per cell, as suggested in prior work 40 ) will be needed to provide more stable reference estimates. Lastly, as this study focused on cognitively healthy older adults, its findings may have limited implications for clinical populations. Further research involving individuals with cognitive impairments will be valuable to explore the assessment's broader clinical utility.
Conclusions
In summary, both age and educational level showed consistent associations with performance across most cognitive domains assessed by Cognosis, whereas memory measures showed limited associations with these demographic factors. Additionally, in certain age and education subgroups, gender differences were observed, with highly educated female participants generally demonstrating superior performance. The reference data presented in this study provide preliminary reference values to support clinical interpretation of this newly developed neuropsychological battery in the Korean elderly population. Furthermore, these data may serve as a valuable resource for future cross-cultural comparisons of the assessment battery.
Supplemental Material
sj-docx-1-dhj-10.1177_20552076261435931 - Supplemental material for Preliminary normative study of Cognosis: A digital cognitive assessment battery for Korean elders
Supplemental material, sj-docx-1-dhj-10.1177_20552076261435931 for Preliminary normative study of Cognosis: A digital cognitive assessment battery for Korean elders by Hyeonjin Kim, Sehee Shim, Sanghoon Lee, Juhye Kim, Soowon Park, Hana Kim, Jun-Young Lee and Jung Hae Youn in DIGITAL HEALTH
Footnotes
Acknowledgements
During the preparation of this work, the authors used ChatGPT to help refine and edit the manuscript. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.
ORCID iDs
Ethical considerations
The study was approved by the Institutional Review Board of Cha University (IRB No. 1044308-202209-HR-053).
Contributorship
Hyeonjin Kim: Formal analysis; Methodology; Visualization; Writing—original draft; and Writing—review & editing.
Sehee Shim, Sanghoon Lee, and Juhye Kim: Conceptualization; Data curation; Investigation; and Software.
Soowon Park: Supervision; Validation; and Writing—review & editing.
Hana Kim, Jun-Young Lee, and Jung Hae Yoon: Conceptualization; Funding acquisition; Project administration; Resources; and Supervision.
Jung Hae Yoon: Writing—review & editing.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by a grant of Korean ARPA-H Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2024-00512374).
Gurantor
The corresponding author is the guarantor of this work and accepts full responsibility for the integrity of the work as a whole.
Informed Consent
Written informed consent was obtained from all participants prior to study participation.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Supplemental material
Supplemental material for this article is available online.