Hostname: page-component-78c5997874-8bhkd Total loading time: 0 Render date: 2024-11-03T08:00:58.579Z Has data issue: false hasContentIssue false

Assessing harmonized intelligence measures in a multinational study

Published online by Cambridge University Press:  23 February 2024

Mariah DeSerisy*
Affiliation:
Columbia University Medical Center, Mailman School of Public Health, Columbia University, New York, NY, USA Columbia University Irving Medical Center, Columbia University, New York, NY, USA
Melanie M. Wall
Affiliation:
Columbia University Irving Medical Center, Columbia University, New York, NY, USA Department of Psychiatry, The New York State Psychiatric Institute, New York, NY, USA
Terry E. Goldberg
Affiliation:
Columbia University Irving Medical Center, Columbia University, New York, NY, USA
Marcelo C. Batistuzzo
Affiliation:
Department of Psychiatry, Faculdade de Medicina, Universidade de São Paulo, São Paulo, Brazil Department of Methods and Techniques in Psychology, Pontifical Catholic University, São Paulo, Brazil
Katherine Keyes
Affiliation:
Columbia University Medical Center, Mailman School of Public Health, Columbia University, New York, NY, USA
Niels T. de Joode
Affiliation:
Department of Psychiatry, Amsterdam UMC, Vrije Universiteit Amsterdam, Amsterdam, Netherlands Department of Anatomy and Neuroscience, Amsterdam UMC, Amsterdam Neuroscience, Vrije Universiteit Amsterdam, Amsterdam, Netherlands
Christine Lochner
Affiliation:
SAMRC Unit on Risk & Resilience in Mental Disorders, Department of Psychiatry, Stellenbosch University, Stellenbosch, South Africa
Clara Marincowitz
Affiliation:
SAMRC Unit on Risk & Resilience in Mental Disorders, Department of Psychiatry, Stellenbosch University, Stellenbosch, South Africa
Madhuri Narayan
Affiliation:
Department of Clinical Psychology, National Institute of Mental Health & Neuro Sciences (NIMHANS), Institute of National Importance (INI), Bangalore, India
Nitin Anand
Affiliation:
Department of Clinical Psychology, National Institute of Mental Health & Neuro Sciences (NIMHANS), Institute of National Importance (INI), Bangalore, India
Amy M. Rapp
Affiliation:
Columbia University Irving Medical Center, Columbia University, New York, NY, USA Department of Psychiatry, The New York State Psychiatric Institute, New York, NY, USA
Dan J. Stein
Affiliation:
SAMRC Unit on Risk & Resilience in Mental Disorders, Department of Psychiatry and Neuroscience Institute, University of Cape Town, Cape Town, South Africa
H. Blair Simpson
Affiliation:
Columbia University Irving Medical Center, Columbia University, New York, NY, USA Department of Psychiatry, The New York State Psychiatric Institute, New York, NY, USA
Amy E. Margolis
Affiliation:
Columbia University Irving Medical Center, Columbia University, New York, NY, USA Department of Psychiatry, The New York State Psychiatric Institute, New York, NY, USA
*
Corresponding author: Mariah DeSerisy; Email: [email protected]
Rights & Permissions [Opens in a new window]

Abstract

Studies examining the neurocognitive and circuit-based etiology of psychiatric illness are moving toward inclusive, global designs. A potential confounding effect of these associations is general intelligence; however, an internationally validated, harmonized intelligence quotient (IQ) measure is not available. We describe the procedures used to measure IQ across a five-site, multinational study and demonstrate the harmonized measure’s cross-site validity. Culturally appropriate intelligence measures were selected: four short-form Wechsler intelligence tests (Brazil, Netherlands, South Africa, United States) and the Binet Kamat (India). Analyses included IQ scores from 255 healthy participants (age 18–50; 42% male). Regression analyses tested between-site differences in IQ scores, as well as expected associations with sociodemographic factors (sex, socioeconomic status, education) to assess validity. Harmonization (e.g., a priori selection of tests) yielded the compatibility of IQ measures. Higher IQ was associated with higher socioeconomic status, suggesting good convergent validity. No association was found between sex and IQ at any site, suggesting good discriminant validity. Associations between higher IQ and higher years of education were found at all sites except the United States. Harmonized IQ scores provide a measure of IQ with evidence of good validity that can be used in neurocognitive and circuit-based studies to control for intelligence across global sites.

Topics structure

Topic(s)

Type
Research Article
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2024. Published by Cambridge University Press

Impact statement

As mental health research shifts toward more inclusive, global studies, there is an increasing need for a harmonized measure of intelligence for use in multinational studies in order to address the potentially confounding effects of intelligence on mental health outcomes. To date, no work has examined the convergent and divergent validity of harmonized intelligence scores across multiple study sites located in different countries. We demonstrated that the full-scale intelligence quotient (IQ) measure harmonized across five multinational sites correlated with socioeconomic status and educational attainment indicating convergent validity and showed that it did not correlate with sex indicating discriminant validity. Site-specific effects were observed and are discussed in the context of their implications for future analyses with combined data across these global sites. The confounding effect of individual differences in intelligence among individuals with neuropsychiatric disorders presents unique challenges for global investigations of mental health across different countries. Our data suggest that this can be mitigated by incorporating a prospectively harmonized valid measure of IQ into analyses to adjust for this confounding, providing preliminary support for using such an approach in future multinational studies.

Introduction

Neuropsychiatric disorders account for as much as 10% of the disease burden worldwide (Santomauro et al., Reference Santomauro, Mantilla Herrera, Shadid and Ferrari2021); however, access to mental health care and research to support such care remains scarce (World Health Organization, Mental Health Determinants and Populations Team, 2001). Studies examining neurocognitive functioning in and neural circuitry of psychiatric illnesses are moving toward more inclusive and global designs. Such work raises the need to address challenges inherent in measuring neurocognitive abilities in different countries that may vary in terms of resources or language, factors known to be associated with performance on cognitive tests.

Intelligence testing is a commonly used tool in research to address individual differences in cognitive capacities across participants by measuring the ability to use information or abstract reasoning to answer questions, make predictions and learn from experience across a number of domains (Deary, Reference Deary2012; Russell, Reference Russell2020). Individual differences in intelligence are important to include in studies designed to measure cognitive problems associated with psychiatric diagnoses because intelligence is associated with psychiatric symptoms and with performance on cognitive tests (Abramovitch et al., Reference Abramovitch, Anholt, Raveh-Gottfried, Hamo and Abramowitz2018; Ruiz et al., Reference Ruiz, Raugh, Bartolomeo and Strauss2020; Thompson et al., Reference Thompson, Babicz, Matchanova and Woods2020a). The potentially confounding effect of individual differences in intelligence on cognitive performance presents unique challenges for global investigations of mental health across different countries especially given that there is no existing best practice for how to measure intelligence across different countries.

Intelligence quotients (IQs) are thought to measure global g, the theorized common factor representing human intelligence (Spearman, Reference Spearman1904). Global g, or g-factor, has been argued to represent a universal human phenomenon (Warne and Burningham, Reference Warne and Burningham2019; Russell, Reference Russell2020); however, the way g manifests is likely to be context specific (i.e., skills useful in an urban context might be different in a rural context) (Warne and Burningham, Reference Warne and Burningham2019; Russell, Reference Russell2020). As such, it is essential to interpret results from intelligence testing within the context of a specific country, region or study site in global studies of mental health.

Although global collaborations examining cognitive outcomes are increasing, the method for handling measures of intelligence has varied widely and has not focused on validity of the measure across sites. The majority of work that has incorporated IQ scores across multiple sites have leveraged a full-scale IQ score regardless of the assessment tool used (e.g., van Bakel et al., Reference van Bakel, Einarsson, Arnaud, Craig, Michelsen, Pildava, Uldall and Cans2014; Sentenac et al., Reference Sentenac, Benhammou, Aden, Ancel, Bakker, Bakoy, Barros, Baumann, Bilsteen, Boerch, Croci, Cuttini, Draper, Halvorsen, Johnson, Källén, Land, Lebeer, Lehtonen, Maier, Marlow, Morgan, Ni, Raikkonen, Rtimi, Sarrechia, Varendi, Vollsaeter, Wolke, Ylijoki and Zeitlin2021) or in big data sets, aggregation is done restricting only to sites that have the same IQ measures (Bedford et al., Reference Bedford, Park, Devenyi, Tullo, Germann, Patel, Anagnostou, Baron-Cohen, Bullmore, Chura, Craig, Ecker, Floris, Holt, Lenroot, Lerch, Lombardo, Murphy, Raznahan, Ruigrok, Smith, Spencer, Suckling, Taylor, Thurm, Lai and Chakravarty2020). Rarely have studies attempted to combine sites with different measures of intelligence from across continents (Mortillo and Mulle, Reference Mortillo and Mulle2021; Wallert et al., Reference Wallert, Rennie, Ferreira, Muehlboeck, Wahlund, Westman and Ekman2021). For example, Mortillo and Mulle (Reference Mortillo and Mulle2021) combined data as well as types of tests across countries by comparing country-specific norm-referenced standard scores and also dichotomizing participants into groups based on intellectual disability status. In contrast, Wallert et al. (Reference Wallert, Rennie, Ferreira, Muehlboeck, Wahlund, Westman and Ekman2021) utilized principal components analysis to extract a g-factor from multiple cognitive tests by combining data from participants in North America and Sweden. None of the studies provided any demonstration of the validity of the intelligence measure across sites. We address this gap by proposing to use prospectively harmonized compatible measures with country-specific norms and then to demonstrate that this harmonized measure shows convergent and discriminant validity across sites. Of note, data harmonization is a tool that can be used to maintain the integrity of context-specific data, such as IQ, while also pooling across contexts to facilitate large-scale global collaborations. Harmonization can be prospective, via careful selection of culturally relevant, reliable and valid measures occurring after data collection has started and leveraging statistical approaches to ensure data compatibility (Griffith et al., Reference Griffith, van den Heuvel, Fortier, Hofer, Raina, Sohel, Payette, Wolfson and Belleville2013, Reference Griffith, van den Heuvel, Raina, Fortier, Sohel, Hofer, Payette, Wolfson, Belleville, Kenny and Doiron2016).

From psychometrics, the validity of a measure is determined via its consistent associations with variables theoretically predicted to be related to it in specific ways, that is, convergent and discriminant validity (Campbell and Fiske, Reference Campbell and Fiske1959). Intelligence is both heritable and malleable (Sauce and Matzel, Reference Sauce and Matzel2018), with strong bidirectional associations with sociodemographic factors, including socioeconomic status (SES, Strenze, Reference Strenze2007) and education (Ritchie and Tucker-Drob, Reference Ritchie and Tucker-Drob2018; Lövdén et al., Reference Lövdén, Fratiglioni, Glymour, Lindenberger and Tucker-Drob2020; Feinkohl et al., Reference Feinkohl, Kozma, Borchers, SJT, Kruppa, Winterer, Spies and Pischon2021). Of note, SES and education have each been shown to differentially associate with verbal (Matarazzo and Herman, Reference Matarazzo and Herman1984; Bornstein et al., Reference Bornstein, Suga and Prifitera1987; Shuttleworth-Edwards et al., Reference Shuttleworth-Edwards, Kemp, Rust, Muirhead, Hartman and Radloff2004; Walker et al., Reference Walker, Batchelor and Shores2009; Chapman et al., Reference Chapman, Fiscella, Duberstein, Kawachi and Muennig2014) and perceptual abilities (Matarazzo and Herman, Reference Matarazzo and Herman1984; Bornstein et al., Reference Bornstein, Suga and Prifitera1987; Shuttleworth-Edwards et al., Reference Shuttleworth-Edwards, Kemp, Rust, Muirhead, Hartman and Radloff2004; Mani et al., Reference Mani, Mullainathan, Shafir and Zhao2013; Piccolo et al., Reference Piccolo, Arteche, Fonseca, Grassi-Oliveira and Salles2016). In contrast, other demographic factors, such as sex, are less correlated with intellectual abilities. Sex differences in FSIQ have not been consistently found (e.g., Colom et al., Reference Colom, García, Juan-Espinosa and Abad2002; Daseking et al., Reference Daseking, Petermann and Waldmann2017; Halpern and Wai, Reference Halpern, Wai and Sternberg2019). However, there is evidence to suggest there may be sex-specific differences in performance on individual subtests or across specific domains (e.g., Irwing, Reference Irwing2012; Pezzuti et al., Reference Pezzuti, Tommasi, Saggino, Dawe and Lauriola2020). In sum, when attempting to confirm the validity of a cross-national IQ measure, we would expect to find positive correlations between the IQ measure, SES and education (convergent validity), but to find minimal or no associations between the IQ measure and sex (discriminant validity).

This manuscript reports on the prospective harmonization process used to select culturally appropriate IQ measures across sites from five countries collected as part of a study examining cognitive and neurobiological correlates of obsessive–compulsive disorder (OCD; Simpson et al., Reference Simpson, van den Heuvel, Miguel, Reddy, Stein, Lewis-Fernández, Shavitt, Lochner, Pouwels, Narayanawamy, Venkatasubramanian, Hezel, Vriend, Batistuzzo, Hoexter, de Joode, Costa, de Mathis, Sheshachala, Narayan, van Balkom, Batelaan, Venkataram, Cherian, Marincowitz, Pannekoek, Stovezky, Mare, Liu, Otaduy, Pastorello, Rao, Katechis, van Meter and Wall2020) compared to healthy participants. We leverage the harmonized intelligence measure obtained from healthy participants to examine the measure’s convergent and discriminant validity across sites in comparison to sociodemographic factors.

Methods

Participants

The parent study recruited and evaluated a large and diverse sample of medication-free adults with OCD and matched healthy participants across five academic medical sites located in Brazil, India, the Netherlands, South Africa and the United States. A full description of the parent study protocol can be found elsewhere (Simpson et al., Reference Simpson, van den Heuvel, Miguel, Reddy, Stein, Lewis-Fernández, Shavitt, Lochner, Pouwels, Narayanawamy, Venkatasubramanian, Hezel, Vriend, Batistuzzo, Hoexter, de Joode, Costa, de Mathis, Sheshachala, Narayan, van Balkom, Batelaan, Venkataram, Cherian, Marincowitz, Pannekoek, Stovezky, Mare, Liu, Otaduy, Pastorello, Rao, Katechis, van Meter and Wall2020). Given our focus on assessing the validity of the IQ measure across sites, we included only healthy control participants. Subjects with OCD may exhibit systematic differences in IQ (Abramovitch et al., Reference Abramovitch, Anholt, Raveh-Gottfried, Hamo and Abramowitz2018) that might be associated with the validity assessment. A total of 256 healthy participants (n = 255 with completed intelligence measure) were recruited across all five sites and selected to match the OCD sample in distribution on age, sex and educational level (within sites but not necessarily between sites). Healthy participants were aged 18–50 years and were not eligible to participate if they had a first-degree relative with OCD or tic disorder, current or past use of psychotropic medications or current or lifetime psychiatric disorder other than major depressive disorder or anxiety disorders (if not in past year). Importantly, healthy participants were also not eligible if they had an FSIQ score below 80.

Prospectively chosen intelligence assessments

Intelligence tests were chosen in consultation with local experts to determine the most context-valid and appropriate test for use at each site, keeping in mind the need for compatibility across sites (Table 1). Thus, intelligence testing was performed using different instruments depending on the site location, local population characteristics and local dominant language. When available, preference was given to short forms of the Wechsler tests to minimize participant burden, reduce cross-site heterogeneity and maximize harmonization opportunities. Of note, discrete ability scores (Perceptual Reasoning Index [PIQ] and Verbal Comprehension Index [VIQ]) were derived whenever possible, as described below.

Table 1. Prospectively chosen intelligence measures across sites

Brazil

The Brazilian site (located in San Paolo) utilized the Brazilian version of the Wechsler Abbreviated Scale of Intelligence, First Edition (WASI-I; Wechsler, Reference Wechsler1999; Trentini et al., Reference Trentini, Yates and Vs2014) administered in Brazilian Portuguese by bachelor’s level psychologist evaluators trained by a post-doctoral level psychologist. The WASI-I consists of Block Design, Matrix Reasoning, Vocabulary and Similarities subtests and derives an examinee’s PIQ, VIQ and FSIQ. Evaluators were trained to reliability and supervised by a doctoral level clinician with >10 years of expertise in neuropsychological assessment. All protocols were scored by the same professional, the supervisor, to ensure ongoing reliability. Tests were scored using publisher norms developed with Brazilian populations (Trentini et al., Reference Trentini, Yates and Vs2014).

India

The India site (located in Bangalore) utilized the Binet Kamat Test of Intelligence (Kamat, Reference Kamat1968) administered in English or Kannada by bilingual evaluators depending on the preference of the participant and based on their language proficiency. Notably, an Indian version of Wechsler tests is not available; therefore, the Binet Kamat was selected as an intelligence test with available local norms. The intelligence measure was administered by master’s level and doctoral-level student clinical psychology evaluators. The Binet Kamat Test includes both verbal and nonverbal items but does not consist of specific subtests or derive subtest scores. Instead, the Binet Kamat derives only an FSIQ score. Evaluators were trained to reliability by doctoral-level clinicians with expertise in neuropsychological assessment and supervised by a doctoral-level clinician. Every fifth test protocol was double-scored by the test administrator and a doctoral-level clinician to ensure ongoing reliability. Tests were scored using norms developed with Indian populations (Kamat, Reference Kamat1968). Despite their age, recent evidence suggests that these norms are still valid among Indian participants (Roopesh, Reference Roopesh2020).

Netherlands

The Netherlands site (located in Amsterdam) utilized four selected subscales from the Netherlands version of the Wechsler Adult Intelligence Scale, Fourth Edition (Wechsler, Reference Wechsler2009) administered in Dutch. Completed subtests included Block Design, Matrix Reasoning, Vocabulary and Similarities to match other sites and derive an examinee’s PIQ, VIQ and FSIQ. Evaluations were completed by doctoral students, master’s students and a bachelor’s level research assistant via iPads. Evaluators were trained to reliability by a doctoral-level clinician with expertise in neuropsychological assessment and supervised by a doctoral-level clinician. Every fifth test protocol was reviewed, and the Vocabulary and Similarities were double-scored by the test administrator and a doctoral-level supervisor to ensure ongoing reliability. Matrix Reasoning and Block Design subsets were automatically generated based on participants’ iPad responses. Tests were scored using publisher norms developed with Dutch and Flemish populations (Wechsler, Reference Wechsler2009).

South Africa

The South Africa site (located in Cape Town) utilized the English version of the WASI, Second Edition (WASI-II; Wechsler, Reference Wechsler2011) administered by bilingual master’s and doctoral-level evaluators. Participants completed the test in either English or Afrikaans, depending on the preference of the participant and based on their language proficiency and the language in which they completed the majority of their education. Of note, an Afrikaans version of Wechsler tests is not available; however, the majority of South Africans in the catchment population were bilingual. Specifically, the majority of participants reported their first language as Afrikaans but performing most educational and occupational duties in English. Test items and directions from the English version of the WASI-II were translated to Afrikaans by bilingual study team members to produce a standardized Afrikaans assessment; of note, test items were directly translated from English to Afrikaans, which may or may not preserve the intended item difficulty. When requested (n = 18), the translated assessment was presented. The WASI-II consists of Block Design, Matrix Reasoning, Vocabulary and Similarities subtests and derives an examinee’s PIQ, VIQ and FSIQ. Evaluators were trained to reliability and supervised by a doctoral-level clinician with expertise in neuropsychological assessment. Every fifth test protocol was reviewed and the Vocabulary and Similarities were double-scored by the test administrator and the doctoral-level supervisor to ensure ongoing reliability. Tests were scored using U.S. publisher norms, as South African norms are not available for the WASI-II. Notably, an alternative test instrument with local norms has not been developed. Cross-site reliability was assessed through monthly meetings with the U.S. site (also utilizing the WASI-II) in which a team of six raters independently rated a test protocol and scores were determined by consensus.

United States

The U.S. site (located in New York City) utilized the WASI-II (Wechsler, Reference Wechsler2011) administered in English. PIQ, VIQ and FSIQ were derived. Evaluators consisted of bachelor’s level research assistants trained to reliability in the administration of the WASI-II. Evaluators were trained by doctoral-level clinicians with expertise in neuropsychological assessment and supervised by a doctoral-level clinician. Every fifth test protocol in entirety was double-scored by the test administrator and a doctoral-level supervisor to ensure ongoing reliability. Tests were scored using publisher norms developed with U.S. populations. As described above, cross-site reliability was assessed through monthly structured meetings with the South Africa site.

Sociodemographic factors for assessing convergent and discriminant validity

Educational attainment, or years of education, is known to be associated with IQ scores and interact with SES (Ritchie and Tucker-Drob, Reference Ritchie and Tucker-Drob2018). Further, it has been used to approximate SES because it can be obtained for all participants, in contrast to other measures such as occupation or income that are associated with family structure (i.e., stay-at-home parents) and retirement age, and is typically robust to late-life health impairments (Liberatos et al., Reference Liberatos, Link and Kelsey1988; Elo and Preston, Reference Elo and Preston1996). In this study, years of education refers to the number of completed (i.e., passed) years of schooling, beginning with the first grade, and was prospectively determined to be a valid harmonized measure across the five countries. Additionally, this method of measuring educational attainment has been used in previous multinational studies (Thompson et al., Reference Thompson, Jahanshad and CRK2020b).

The WAMI Index (Psaki et al., Reference Psaki, Seidman, Miller and Investigators2014) measures access to resources and living conditions that differ between developing and developed countries, such as access to improved water/sanitation, assets (e.g., housing resources), maternal education and income. The WAMI provides a summary index score as well as section scores examining 1) water/sanitation; 2) assets (e.g., possessions, number of rooms in the family home); 3) maternal educational attainment and 4) household income in local currency. This measure was prospectively chosen as it has been shown to validly measure SES across different countries (Psaki et al., Reference Psaki, Seidman, Miller and Investigators2014; Pradhan et al., Reference Pradhan, Ali, Hasnani, Bhamani and Karmaliani2018).

Sex was determined by the participant’s self-report.

Data analytic plan

Descriptive summaries (means, standard deviations) and one-way analysis of variance or chi-squared (χ 2) tests were used to assess differences between sites in participant sociodemographic characteristics and IQ measures: FSIQ, VIQ and PIQ. Scheffé post hoc tests were used to test pairwise mean differences. To assess the construct validity (Terwee et al., Reference Terwee, Bot, de Boer, DAWM, Knol, Dekker, Bouter and de Vet2007; Mokkink et al., Reference Mokkink, Terwee, Knol, Stratford, Alonso, Patrick, Bouter and de Vet2010), we hypothesized that FSIQ, VIQ and PIQ will all correlate positively with years of education and WAMI (convergent validity) and will not correlate with sex (discriminant validity). Further, we hypothesized these associations to be found across sites and within each site. To test these validity hypotheses, we used general linear models (GLMs) for each IQ measure as the outcome predicted by site, sex, education and WAMI and included an interaction of site with each of the four sociodemographic measures to test the similarity of associations across sites. Given the known associations between age and IQ (i.e., declines in processing speed and fluid reasoning beginning in early adulthood and becoming impairing in elderly individuals age > 75; Miller et al., Reference Miller, Myers, Prinzi and Mittenberg2009; Baxendale, Reference Baxendale2011; Singh-Manoux et al., Reference Singh-Manoux, Kivimaki, Glymour, Elbaz, Berr, Ebmeier, Ferrie and Dugravot2012; Kremen et al., Reference Kremen, Moore, Franz, Panizzon, Lyons, Finkel and Reynolds2014), we include age and age by site interactions in our models to control for potential confounding by age. We note that IQ test norms adjust for age (Wechsler, Reference Wechsler2009), but we include age in our models to account for differences in ages across sites. The India site was excluded from the analyses of VIQ and PIQ as these subtests were not available. Effect sizes were determined using partial eta-squared (η 2). Common rules of thumb for qualifying the size of partial η 2 are that 0.01 is small, 0.06 is medium and 0.14 is large (Richardson, Reference Richardson2011). Analyses were performed using IBM SPSS Statistics version 28 (IBM Corp., Armonk, NY, United States) and alpha was set at p < 0.05 (two-tailed) for all analyses.

Many participants across the global sites were multilingual. Sensitivity analyses evaluated the effects of language proficiency as well as task and task administration on primary results. Language proficiency was determined by asking participants’ their preferred language and determining if that language matched the administration language. This dichotomous variable was then included in the GLMs described above and tested. Because the India site used the Binet Kamat and because task administration was nonstandard in the South Africa site, the primary analysis was conducted without including data from the South Africa or India sites.

Results

Participants

Similar numbers of healthy participants were recruited at all five sites (range, n = 50–53), with average age across sites ranging from 27.7 to 32.7 and the gender distribution from 35% to 54% male (Table 2). Supplementary Table S1 describes participants’ ethno-racial backgrounds in detail. Across the sites, years of education were higher than the general population of the world (8.0–8.7 years; “Average years of schooling”, n.d.; Barro and Lee, Reference Barro and Lee2013) and of each country (“Average years of schooling”, n.d.), with the highest level in Brazil (17 years, country average = 8 years; “Average years of schooling”, n.d.) and lowest (though still high) in South Africa (15 years, country average = 13 years; “Average years of schooling”, n.d.) likely due to convenience sampling occurring at academic research institutions. WAMI Index scores were also higher than the general population (0.58, Psaki et al., Reference Psaki, Seidman, Miller and Investigators2014), highest in the United States (0.83) and Netherlands (0.82) and lowest in India (0.68), albeit still higher than the general population.

Table 2. Sociodemographic characteristics of healthy adult participants across sites

Abbreviations: χ2, chi-square; F, ANOVA F statistic; SD, standard deviation; SES, socioeconomic status derived from WAMI.

a Sites shown with = indicate they were not significantly different at p < 0.05. Sites shown with > indicate significant difference at p < 0.05.

b One participant at South Africa did not have intelligence scores and was dropped from further analyses.

Summary statistics for IQ measures across sites

The distribution of raw FSIQ scores at each site fell within the expected ranges (Supplementary Figure S1). Means of FSIQ, VIQ and PIQ (Table 3) were generally higher at every site than the standard population mean of 100 but had standard deviations ranging from 10.7 to 12.9 as expected. There were no differences between IQ indices between sites when controlling for site differences in demographics (Table 3, for raw data in Table 3: Supplement, Supplementary Table S2 and Supplementary Figure S1) (FSIQ: p = 0.46, η2 = 0.016; VIQ: p = 0.54, η2 = 0.012 and PIQ: p = 0.20, η2 = 0.025).

Table 3. Mean intelligence scores across sites controlling for biological sex, years of education and SES

Abbreviations: F, ANOVA F statistic; FSIQ, full-scale intelligence quotient; PIQ, perceptual intelligence quotient; VIQ, verbal intelligence quotient.

Note: Unadjusted results in the Supplementary Table S2.

* The Binet Kamat Test does not provide index scores for verbal or perceptual reasoning.

Convergent and discriminant validity

Consistent with the convergent validity hypotheses, we find higher SES, as measured by the WAMI index, was significantly associated with increased FSIQ scores (F(1,230) = 12.48, p < 0.001; partial η2 = 0.051) and this was consistent by site as shown by the lack of interaction and respective small effect size of SES by site (F(4,222) = 0.61, p = 0.66, partial η2 = 0.011). Specifically, FSIQ increased 0.32 points for every 1 standard deviation (i.e., 0.10 point) increase in WAMI index score (Figure 1A and Table 4).

Figure 1. Associations between FSIQ scores and sociodemographics. (A) Main effect of WAMI index score (SES) was significant. (B) Full-scale IQ score is positively associated with educational attainment in Brazil (blue), India (maroon), Netherlands (teal) and South Africa (purple) and negatively associated with educational attainment in the United States (green). Main effects of (C) sex were not significant. Figures depicting individual sites in the Supplementary Figure S2. SES, socioeconomic status from the WAMI; FSIQ, full-scale IQ.

Table 4. ANOVA predicting FSIQ scores

Abbreviations: df, degrees of freedom; SES, socioeconomic status from the WAMI; SS, type III sum of squares.

Adjusted R 2 = 0.23.

The results of the convergent validity hypothesis for years of education were mixed because the effect of education on FSIQ was found to differ significantly by site (F(4,230) = 3.42, p = 0.01, partial η2 = 0.056), such that each country showed a positive association between years of education and FSIQ, except the United States. Specifically, FSIQ increased 9.45 points in Brazil, 4.35 points in India, 10.73 points in the Netherlands and 4.52 points in South Africa for every standard deviation (2.5-year) increase in education; however, in the United States, FSIQ decreased by 13.52 points for every 2.5-year increase in education (Figure 1B).

Consistent with the discriminant validity hypotheses, we found that (F(1,230) = 0.21, p = 0.64, η2 = 0.001) sex (F(1,230) = 2.97, p = 0.09, η2 = 0.013) was not associated with FSIQ (Figure 1C). Also, results for convergent and discriminant validity hypotheses were consistent for both VIQ and PIQ (Supplementary Table S3 and Supplementary Figures S2 and S3). Finally, sensitivity analyses including additional control for a measure of language proficiency did not change this pattern of results and was itself not found to be a significant predictor of FSIQ, VIQ or PIQ (Supplementary Table S4). Finally, excluding data from the South Africa and India sites did not alter the findings (Supplementary Table S5).

Discussion

Herein, we described the collection of harmonized IQ data for use in a large-scale, multisite, global study. Researchers conducting this study performed considerable prospective harmonization procedures prior to the onset of data collection to ensure the compatibility of IQ scores across sites. Prospective harmonization included consultation with local experts and attempts to utilize a single family of tests (i.e., Wechsler tests) in as many sites as possible, with the goal of yielding largely compatible measures with country-specific norms. Consistent with our discriminant validity hypotheses, associations between sex and IQ were not detected in this healthy participant sample. Consistent with our convergent validity hypothesis, higher FSIQ, as well as VIQ and PIQ, were associated with higher SES across the entire sample. The hypothesized positive association with education was confirmed in four of the five sites but did not hold for the United States. Validation of the prospectively harmonized IQ measure developed in this study provides preliminary support for using such an approach in future studies.

SES is known to be closely tied to socioenvironmental improvements, and correlates of lower SES such as lack of access to clean water/sanitation (Dearden et al., Reference Dearden, Brennan, Behrman, Schott, Crookston, Humphries, Penny and Fernald2017; Orgill-Meyer and Pattanayak, Reference Orgill-Meyer and Pattanayak2020) fewer household assets or resources (Hackman et al., Reference Hackman, Farah and Meaney2010; Barreto et al., Reference Barreto, Sánchez de Miguel, Ibarluzea, Andiarena and Arranz2017; Flensborg-Madsen et al., Reference Flensborg-Madsen, Hanne-Lise Falgreen and Mortensen2020; Zhang, Reference Zhang2021), and lower maternal educational attainment (Lawlor et al., Reference Lawlor, Najman, Batty, O’Callaghan, Williams and Bor2006; Crookston et al., Reference Crookston, Forste, McClellan, Georgiadis and Heaton2014; Lewinn et al., Reference Lewinn, Bush, Batra, Tylavsky and Rehkopf2020) are closely tied to lower IQ scores in previous global cohort studies of children and youth. In our global study that collected data from healthy adult participants in five sites spanning five continents, IQ increased an average of three points for every 0.1-point increase in SES (WAMI index score) across these five study sites. This finding that higher SES was associated with higher IQ across sites adds to the knowledge base by showing convergently valid, stable associations between these factors across a broad range of SES indicators in a multinational context. Moreover, our finding that IQ scores did not associate with sex is also consistent with prior findings (Colom et al., Reference Colom, García, Juan-Espinosa and Abad2002; Miller et al., Reference Miller, Myers, Prinzi and Mittenberg2009; Baxendale, Reference Baxendale2011; Daseking et al., Reference Daseking, Petermann and Waldmann2017; Halpern and Wai Reference Halpern, Wai and Sternberg2019; Pezzuti et al., Reference Pezzuti, Tommasi, Saggino, Dawe and Lauriola2020) and contributes to the confidence that the procedural harmonization across sites did not introduce any type of systematic bias related to participant characteristics. Given that sex (Weber et al., Reference Weber, Gupta, Abdalla, Cislaghi, Meausoone and Darmstadt2021) could introduce bias in global health studies, our finding that IQ was not associated with sex suggests that our data are robust to these potential demographic biases.

Prospective harmonization yielded strong data compatibility in IQ but is not a panacea. As seen in our results, other procedures still may be needed to control differences in IQ across sites. Specifically, we found that associations between education and IQ (FSIQ, VIQ) varied by site such that higher levels of education were positively associated with IQ at every site except the United States, indicating site-specific associations (Teasdale and Owen, Reference Teasdale and Owen2005; Dutton et al., Reference Dutton, van der Linden and Lynn2016; Bratsberg and Rogeberg, Reference Bratsberg and Rogeberg2018; Acosta et al., Reference Acosta, Smith and Kreinovich2019) between education and IQ measures that could not be controlled by prospective harmonization. There are several possible interpretations for our finding of a site by education interaction effect. First, these findings may suggest that increasing years of education in the United States (beyond 12 years of compulsory education, e.g., community college) may not be as associated with increasing IQ scores as they are in other countries. Alternatively, given the nature of our convenience sample, it is likely that our participants are not representative of the U.S. population. Supporting this, previous research has reported associations between educational attainment, IQ and SES in U.S. samples similar to those observed at our Brazil, Netherlands, South Africa and India sites (Ritchie and Tucker-Drob, Reference Ritchie and Tucker-Drob2018). As such, our finding of a site by education interaction effect on IQ scores warrants further investigation in larger samples of more diverse participants in the United States.

Our study also examined expected associations with discrete ability across sites. Convergent and discriminant validity hypotheses were confirmed for both VIQ and PIQ consistent with prior studies (Mascie-Taylor and Gibson, Reference Mascie-Taylor and Gibson1978; Reynolds et al., Reference Reynolds, Chastain, Kaufman and McLean1987; Shuttleworth-Edwards et al., Reference Shuttleworth-Edwards, Kemp, Rust, Muirhead, Hartman and Radloff2004; Mani et al., Reference Mani, Mullainathan, Shafir and Zhao2013).

This study had particular strengths in its prospective harmonization process and large-scale, multinational research design. We were able to leverage measures of both personal educational attainment and a globally sensitive measure of SES to examine the effects of these variables across sites and on FSIQ, VIQ and PIQ. At the same time, the study also had limitations. First, this study consisted of primarily convenience sample, including participants who responded to advertisements and were willing to volunteer to contribute to research as healthy individuals. We acknowledge that this limits the generalizability of our findings to those individuals with both the means and ability to present to multiple study visits and participate in all aspects of a study. In a cross-national context, this becomes even more salient as some participants may be unintentionally excluded due to lack of adequate time or transportation or mistrust in research programs. Future examination of the validity of the IQ score harmonization would benefit from more participants enrolled from wider catchment areas with potentially more study sites within countries that are housed in rural areas or off university campuses. Second, VIQ and PIQ were not available for the India site due to differential IQ assessment procedures. Third, evaluators at different sites had different levels of experience and training in the provision of IQ assessments, which could have influenced our findings. However, influences of differential training were mitigated by rigorous data-checking procedures occurring both within and between study sites and over the course of the study. Our study was not able to account for nuanced differences in language proficiency that may have influenced performance on IQ measures. However, sensitivity analyses examining basic language proficiency in the test language were performed and did not influence our results. Future studies should examine the influence of language proficiency in more detail as it may be associated with cross-national harmonization. Also, our study did not include participants with FSIQ less than 80; and therefore, we cannot presume that our findings are generalizable to individuals across the lower end of the IQ spectrum. Future studies would benefit from inclusion of these individuals to better understand how sociodemographic variables do or do not associate with IQ among the intellectually challenged. Finally, local norms were used in Brazil and India whereas publisher norms were used in the United States, South Africa and the Netherlands. Further, publisher norms standardized in the United States were used in South Africa because no local norms were available. We acknowledge that the use of local norms can substantially influence intelligence scores when compared to using publisher norms with the same population (Duggan et al., Reference Duggan, Awakon, Loaiza and Garcia-Barrera2019) and that local norms may not accurately reflect broader population demographics in the same way as publisher norms (Fernández and Abe, Reference Fernández and Abe2018). However, sensitivity analyses excluding the South Africa site were performed and did not influence our results. Future studies should examine if validity statistics change when using publisher norms (versus local norms) for the purposes of harmonization.

Conclusions

This study examined a harmonized measure of intelligence for use in a large, multinational study. Both convergent and discriminant validity of the IQ score with demographic variables were demonstrated. Our study provides preliminary support that prospective harmonization methods are effective in addressing data compatibility across multinational sites. This validated prospective harmonization offers future studies a blueprint for developing harmonizable, culturally relevant assessment tools across global study sites.

Open peer review

To view the open peer review materials for this article, please visit http://doi.org/10.1017/gmh.2024.22.

Supplementary material

The supplementary material for this article can be found at https://doi.org/10.1017/gmh.2024.22.

Data availability statement

Data are available upon written request.

Acknowledgments

The authors would like to thank the participants for their time, energy and dedication in completing the study procedures. The authors would also like to thank their research assistants and test administrators for all of their time and talent: Gaironessa Hendrics, M.A. (South Africa); Nienke Pannekoek, Ph.D. (South Africa); Lian Taljaard, M.A. (South Africa); Loche Manuel, B.A. (South Africa); Jamila Rocha, B.A. (Brazil); Deise Palermo Ruiz, B.A. (Brazil); Eline Vester, BSc. (Netherlands); Minne Scheper, MSc. (Netherlands); Britt Mestdagh, MSc. (Netherlands); Iza Cools, BSc. (Netherlands); Mahashweta Bhattacharya, Mphil. (India); Yael Stovetzky, B.S. (USA); Rachel Middleton, B.A. (USA); Gabrielle Messner, B.A. (USA) and Sarah Rose, B.A. (USA).

Author contribution

M.D., M.W. and A.E.M. were primarily responsible for writing and editing. M.D. and M.W. were responsible for data analysis. T.E.G., M.C.B., N.T.J., C.L., C.M., M.N., N.A., A.M.R., D.J.S. and H.B.S. were responsible for data acquisition and quality control at their respective sites as well as provided editorial feedback. K.K. provided material statistical support.

Financial support

This paper uses data from a NIMH funded study (R01 MH113250) that is a collaboration between five global sites (sites [Principal Investigators]: Brazil [Drs. Euripedes Miguel and Roseli G. Shavitt]; India [Dr. Janardhan Reddy YC]; Netherlands [Dr. Odile A. van den Heuvel]; South Africa [Drs. Dan J. Stein and Christine Lochner] and USA [Drs. Helen Blair Simpson and Melanie Wall]). M.D. was supported by T32-ES-023772 to Pam Factor-Litvak, Ph.D. and Jeffery Shaman, Ph.D., as well as R01ES032296 to A.E.M. A.M.R. was supported by T32MH015144 to Steven Roose, M.D.

Competing interest

In the last 12 months, H.B.S. has received royalties from UpToDate Inc. and a stipend from the American Medical Association for serving as Associate Editor of JAMA Psychiatry. D.J.S. has received consultancy honoraria from Discovery Vitality, Johnson & Johnson, Kanna, L’Oreal, Lundbeck, Orion, Sanofi, Servier, Takeda and Vistagen. The remaining authors have no competing interests to declare.

References

Abramovitch, A, Anholt, G, Raveh-Gottfried, S, Hamo, N and Abramowitz, JS (2018) Meta-analysis of intelligence quotient (IQ) in obsessive-compulsive disorder. Neuropsychology Review 28(1), 111120.CrossRefGoogle ScholarPubMed
Acosta, G, Smith, E and Kreinovich, V (2019) Why IQ test scores are slightly decreasing: Possible system-based explanation for the reversed Flynn effect. Departmental Technical Reports (CS). El Paso, El Paso: The University of Texas. 1342.Google Scholar
Average years of schooling (n.d.) Retrieved March 31, 2023, from https://www.worldeconomics.com/Indicator-Data/ESG/Social/Mean-Years-of-Schooling/.Google Scholar
Barreto, FB, Sánchez de Miguel, M, Ibarluzea, J, Andiarena, A and Arranz, E (2017) Family context and cognitive development in early childhood: A longitudinal study. Intelligence 65, 1122.CrossRefGoogle Scholar
Barro, RJ and Lee, JW (2013) A new data set of educational attainment in the world, 1950–2010. Journal of Development Economics 104, 184198.CrossRefGoogle Scholar
Batistuzzo, MC, Sheshachala, K, Alschuler, DM, Hezel, DM, Lewis-Fernández, R, de Joode, NT, Vriend, C, Lempert, KM, Narayan, M, Marincowitz, C, Lochner, C, Stein, DJ, Narayanaswamy, JC, Heuvel, OA, Simpson, HB and Wall, M (2023) Cross-national harmonization of neurocognitive assessment across five sites in a global study. Neuropsychology 37(3), 284300.Google Scholar
Baxendale, S (2011) IQ and ability across the adult life span. Applied Neuropsychology 18(3), 164167.Google Scholar
Bedford, SA, Park, MTM, Devenyi, GA, Tullo, S, Germann, J, Patel, R, Anagnostou, E, Baron-Cohen, S, Bullmore, ET, Chura, LR, Craig, MC, Ecker, C, Floris, DL, Holt, RJ, Lenroot, R, Lerch, JP, Lombardo, MV, Murphy, DGM, Raznahan, A, Ruigrok, ANV, Smith, E, Spencer, MD, Suckling, J, Taylor, MJ, Thurm, A MRC AIMS Consortium,Lai, MC and Chakravarty, MM (2020) Large-scale analyses of the relationship between sex, age and intelligence quotient heterogeneity and cortical morphometry in autism spectrum disorder. Molecular Psychiatry 25(3), 614628.Google Scholar
Bornstein, RA, Suga, L and Prifitera, A (1987) Incidence of verbal IQ‐performance IQ discrepancies at various levels of education. Journal of Clinical Psychology 43(3), 387389.Google Scholar
Bratsberg, B and Rogeberg, O (2018) Flynn effect and its reversal are both environmentally caused. Proceedings of the National Academy of Sciences 115(26), 6674.Google Scholar
Campbell, DT and Fiske, DW (1959) Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin 56(2), 81105.Google Scholar
Chapman, B, Fiscella, K, Duberstein, P, Kawachi, I and Muennig, P (2014) Measurement confounding affects the extent to which verbal IQ explains social gradients in mortality. Journal of Epidemiology and Community Health 68(8), 728733.Google Scholar
Colom, R, García, LF, Juan-Espinosa, M and Abad, FJ (2002) Null sex differences in general intelligence: Evidence from the WAIS-III. Spanish Journal of Psychology 5(1), 2935.Google Scholar
Crookston, BT, Forste, R, McClellan, C, Georgiadis, A and Heaton, TB (2014) Factors associated with cognitive achievement in late childhood and adolescence: The young lives cohort study of children in Ethiopia, India, Peru, and Vietnam. BMC Pediatrics 14, 253.Google Scholar
Daseking, M, Petermann, F and Waldmann, H-C (2017) Sex differences in cognitive abilities: Analyses for the German WAIS-IV. Personality and Individual Differences 114, 145150.Google Scholar
Dearden, KA, Brennan, AT, Behrman, JR, Schott, W, Crookston, BT, Humphries, DL, Penny, ME and Fernald, LCH (2017) Does household access to improved water and sanitation in infancy and childhood predict better vocabulary test performance in Ethiopian, Indian, Peruvian and Vietnamese cohort studies? BMJ Open 7(3), e013201e013201.CrossRefGoogle ScholarPubMed
Deary, IJ (2012) Intelligence. Annual Review of Psychology 63, 453482.CrossRefGoogle ScholarPubMed
Duggan, EC, Awakon, LM, Loaiza, CC and Garcia-Barrera, MA (2019) Contributing towards a cultural neuropsychology assessment decision-making framework: Comparison of WAIS-IV norms from Colombia, Chile, Mexico, Spain, United States, and Canada. Archives of Clinical Neuropsychology 34(5), 657681.Google Scholar
Dutton, E, van der Linden, D and Lynn, R (2016) The negative Flynn effect: A systematic literature review. Intelligence 59, 163169.CrossRefGoogle Scholar
Elo, IT and Preston, SH (1996) Educational differentials in mortality: United States, 1979-85. Social Science & Medicine 42(1), 4757.CrossRefGoogle ScholarPubMed
Feinkohl, I, Kozma, P, Borchers, F, SJT, Montfort, Kruppa, J, Winterer, G, Spies, C and Pischon, T (2021) Contribution of IQ in young adulthood to the associations of education and occupation with cognitive ability in older age. BMC Geriatrics 21(1), 346346.Google Scholar
Fernández, AL and Abe, J (2018) Bias in cross-cultural neuropsychological testing: Problems and possible solutions. Culture and Brain 6(1), 135.Google Scholar
Flensborg-Madsen, T, Hanne-Lise Falgreen, E and Mortensen, EL (2020) Early life predictors of intelligence in young adulthood and middle age. PLoS One 15(1), e0228144. https://doi.org/10.1371/journal.pone.0228144Google Scholar
Griffith, LE, van den Heuvel, E, Fortier, I, Hofer, S, Raina, P, Sohel, N, Payette, H, Wolfson, C and Belleville, S (2013) Harmonization of Cognitive Measures in Individual Participant Data and Aggregate Data Meta-Analysis. Rockville, MD: Agency for Healthcare Research and Quality (US).Google Scholar
Griffith, LE, van den Heuvel, E, Raina, P, Fortier, I, Sohel, N, Hofer, SM, Payette, H, Wolfson, C, Belleville, S, Kenny, M and Doiron, D (2016) Comparison of standardization methods for the harmonization of phenotype data: An application to cognitive measures. American Journal of Epidemiology 184(10), 770778.Google Scholar
Hackman, DA, Farah, MJ and Meaney, MJ (2010) Socioeconomic status and the brain: Mechanistic insights from human and animal research. Nature Reviews. Neuroscience 11(9), 651659.CrossRefGoogle ScholarPubMed
Halpern, DF, and Wai, J (2020) Sex differences in intelligence. In Sternberg, RJ (ed.), The Cambridge Handbook of Intelligence. Cambridge Handbooks in Psychology. Cambridge: Cambridge University Press, pp. 317345.Google Scholar
Irwing, P (2012) Sex differences in g: An analysis of the US standardization sample of the WAIS-III. Personality and Individual Differences 53(2), 126131.Google Scholar
Kamat, VV (1968) Measuring intelligence of Indian children. British Journal of Educational Studies 16(2), 225225.Google Scholar
Kremen, WS, Moore, CS, Franz, CE, Panizzon, MS and Lyons, MJ (2014) Cognition in middle adulthood. In Finkel, D and Reynolds, CA (eds), Behavior Genetics of Cognition across the Lifespan. New York: Springer, pp. 105134.Google Scholar
Lawlor, D, Najman, J, Batty, G, O’Callaghan, M, Williams, G and Bor, W (2006) Early life predictors of childhood intelligence: Findings from the mater-university study of pregnancy and its outcomes. Paediatric and Perinatal Epidemiology 20, 148162.Google Scholar
Lewinn, KZ, Bush, NR, Batra, A, Tylavsky, F and Rehkopf, D (2020) Identification of modifiable social and behavioral factors associated with childhood cognitive performance. JAMA Pediatrics 174(11), 10631072.Google Scholar
Liberatos, P, Link, BG and Kelsey, JL (1988) The measurement of social class in epidemiology. Epidemiologic Reviews 10, 87121.CrossRefGoogle ScholarPubMed
Lövdén, M, Fratiglioni, L, Glymour, MM, Lindenberger, U and Tucker-Drob, EM (2020) Education and cognitive functioning across the life span. Psychological Science in the Public Interest 21(1), 641.Google Scholar
Mani, A, Mullainathan, S, Shafir, E and Zhao, J (2013) Poverty impedes cognitive function. Science 341(6149), 976.Google Scholar
Mascie-Taylor, CGN and Gibson, JB (1978) Social mobility and IQ components. Journal of Biosocial Science 10(3), 263276.Google Scholar
Matarazzo, JD and Herman, DO (1984) Relationship of education and IQ in the WAIS--R standardization sample. Journal of Consulting and Clinical Psychology 52(4), 631634.Google Scholar
Miller, LJ, Myers, A, Prinzi, L and Mittenberg, W (2009) Changes in intellectual functioning associated with normal aging. Archives of Clinical Neuropsychology 24(7), 681688.Google Scholar
Mokkink, LB, Terwee, CB, Knol, DL, Stratford, PW, Alonso, J, Patrick, DL, Bouter, LM and de Vet, HCW (2010) The COSMIN checklist for evaluating the methodological quality of studies on measurement properties: A clarification of its content. BMC Medical Research Methodology 10(1), 22.Google Scholar
Mortillo, M and Mulle, JG (2021) A cross-comparison of cognitive ability across 8 genomic disorders. Current Opinion in Genetics & Development 68, 106116.Google Scholar
Orgill-Meyer, J and Pattanayak, SK (2020) Improved sanitation increases long-term cognitive test scores. World Development 132, 104975.Google Scholar
Pezzuti, L, Tommasi, M, Saggino, A, Dawe, J and Lauriola, M (2020) Gender differences and measurement bias in the assessment of adult intelligence: Evidence from the Italian WAIS-IV and WAIS-R standardizations. Intelligence 79, 101436.Google Scholar
Piccolo, LdR, Arteche, AX, Fonseca, RP, Grassi-Oliveira, R and Salles, JF (2016) Influence of family socioeconomic status on IQ, language, memory and executive functions of Brazilian children. Psicologia: Reflexão e Crítica 29(1), 23.Google Scholar
Pradhan, NA, Ali, TS, Hasnani, FB, Bhamani, SS and Karmaliani, R (2018) Measuring socio-economic status of an urban squatter settlement in Pakistan using WAMI index. Journal of the Pakistan Medical Association 68(5), 709714.Google Scholar
Psaki, SR, Seidman, JC, Miller, M and Investigators, Mal-Ed Network (2014) Measuring socioeconomic status in multicountry studies: Results from the eight-country MAL-ED study. Population Health Metrics 12(1), 8.Google Scholar
Reynolds, CR, Chastain, RL, Kaufman, AS and McLean, JE (1987) Demographic characteristics and IQ among adults: Analysis of the WAIS-R standardization sample as a function of the stratification variables. Journal of School Psychology 25(4), 323342.Google Scholar
Richardson, JTE (2011) Eta squared and partial eta squared as measures of effect size in educational research. Educational Research Review 6(2), 135147.Google Scholar
Ritchie, SJ and Tucker-Drob, EM (2018) How much does education improve intelligence? A meta-analysis. Psychological Science 29(8), 13581369.Google Scholar
Roopesh, BN (2020) Binet Kamat Test of intelligence: Administration, scoring and interpretation–an in-depth appraisal. American Indian and Alaska Native Mental Health Research: Journal of the National Center 7(3), 180201.Google Scholar
Ruiz, I, Raugh, IM, Bartolomeo, LA and Strauss, GP (2020) A meta-analysis of neuropsychological effort test performance in psychotic disorders. Neuropsychology Review 30(3), 407424.Google Scholar
Russell, TW (2020) In the Know: Debunking 35 Myths about Human Intelligence. New York: Cambridge University Press.Google Scholar
Santomauro, DF, Mantilla Herrera, AM, Shadid, J, … Ferrari, AJ (2021) Global prevalence and burden of depressive and anxiety disorders in 204 countries and territories in 2020 due to the COVID-19 pandemic. Lancet 398(10312), 17001712.Google Scholar
Sauce, B and Matzel, LD (2018) The paradox of intelligence: Heritability and malleability coexist in hidden gene-environment interplay. Psychological Bulletin 144(1), 2647.CrossRefGoogle ScholarPubMed
Sentenac, M, Benhammou, V, Aden, U, Ancel, P-Y, Bakker, LA, Bakoy, H, Barros, H, Baumann, N, Bilsteen, JF, Boerch, K, Croci, I, Cuttini, M, Draper, E, Halvorsen, T, Johnson, S, Källén, K, Land, T, Lebeer, J, Lehtonen, L, Maier, RF, Marlow, N, Morgan, A, Ni, Y, Raikkonen, K, Rtimi, A, Sarrechia, I, Varendi, H, Vollsaeter, M, Wolke, D, Ylijoki, M and Zeitlin, J (2021) Maternal education and cognitive development in 15 European very-preterm birth cohorts from the RECAP preterm platform. International Journal of Epidemiology 50(6), 18241839. https://doi.org/10.1093/ije/dyab170.Google Scholar
Shuttleworth-Edwards, A, Kemp, R, Rust, A, Muirhead, JL, Hartman, N and Radloff, S (2004) Cross-cultural effects on IQ test performance: A review and preliminary normative indications on WAIS-III test performance. Journal of Clinical and Experimental Neuropsychology 26(7), 903920.Google Scholar
Simpson, HB, van den Heuvel, OA, Miguel, EC, Reddy, YCJ, Stein, DJ, Lewis-Fernández, R, Shavitt, RG, Lochner, C, Pouwels, PJW, Narayanawamy, JC, Venkatasubramanian, G, Hezel, DM, Vriend, C, Batistuzzo, MC, Hoexter, MQ, de Joode, NT, Costa, DL, de Mathis, MA, Sheshachala, K, Narayan, M, van Balkom, AJLM, Batelaan, NM, Venkataram, S, Cherian, A, Marincowitz, C, Pannekoek, N, Stovezky, YR, Mare, K, Liu, F, Otaduy, MCG, Pastorello, B, Rao, R, Katechis, M, van Meter, P and Wall, M (2020) Toward identifying reproducible brain signatures of obsessive-compulsive profiles: Rationale and methods for a new global initiative. BMC Psychiatry 20(1), 68.Google Scholar
Singh-Manoux, A, Kivimaki, M, Glymour, MM, Elbaz, A, Berr, C, Ebmeier, KP, Ferrie, JE and Dugravot, A (2012) Timing of onset of cognitive decline: Results from Whitehall II prospective cohort study. BMJ 344(jan04 4), d7622.Google Scholar
Spearman, C (1904) General intelligence objectively determined and measured. American Journal of Psychology 15, 201293.Google Scholar
Strenze, T (2007) Intelligence and socioeconomic success: A meta-analytic review of longitudinal research. Intelligence 35(5), 401426.Google Scholar
Teasdale, TW and Owen, DR (2005) A long-term rise and recent decline in intelligence test performance: The Flynn effect in reverse. Personality and Individual Differences 39(4), 837843.Google Scholar
Terwee, CB, Bot, SDM, de Boer, MR, DAWM, Windt, Knol, DL, Dekker, J, Bouter, LM and de Vet, HCW (2007) Quality criteria were proposed for measurement properties of health status questionnaires. Journal of Clinical Epidemiology 60(1), 3442.Google Scholar
Thompson, JL, Babicz, MA, Matchanova, A and Woods, SP (2020a) Is HIV disease associated with a discrepancy between premorbid verbal IQ and neurocognitive functions? Journal of Clinical and Experimental Neuropsychology 42(8), 857866.Google Scholar
Thompson, PM, Jahanshad, N, CRK, Ching and ENIGMA Consortium (2020b) ENIGMA and global neuroscience: A decade of large-scale studies of the brain in health and disease across more than 40 countries. Translational Psychiatry 10(1), 100.Google Scholar
Trentini, CM, Yates, DB and Vs, H (2014) Escala de Inteligência Wechsler Abreviada (WASI): Manual Profissional. São Paulo: Pearson.Google Scholar
van Bakel, M, Einarsson, I, Arnaud, C, Craig, S, Michelsen, SI, Pildava, S, Uldall, P and Cans, C (2014) Monitoring the prevalence of severe intellectual disability in children across Europe: Feasibility of a common database. Developmental Medicine and Child Neurology 56(4), 361369.Google Scholar
Walker, AJ, Batchelor, J and Shores, A (2009) Effects of education and cultural background on performance on WAIS-III, WMS-III, WAIS-R and WMS-R measures: Systematic review. Australian Psychologist 44(4), 216223.Google Scholar
Wallert, J, Rennie, A, Ferreira, D, Muehlboeck, JS, Wahlund, LO, Westman, E, Ekman, U and ADNI Consortium, and MemClin Steering Committee (2021) Cognitive dedifferentiation as a function of cognitive impairment in the ADNI and MemClin cohorts. Aging 13(10), 1343013442.CrossRefGoogle Scholar
Warne, RT and Burningham, C (2019) Spearman’s g found in 31 non-Western nations: Strong evidence that g is a universal phenomenon. Psychological Bulletin 145(3), 237272.Google Scholar
Weber, AM, Gupta, R, Abdalla, S, Cislaghi, B, Meausoone, V and Darmstadt, GL (2021) Gender-related data missingness, imbalance and bias in global health surveys. BMJ Global Health 6(11), e007405.Google Scholar
Wechsler, D (1999) Wechsler Abbreviated Scale of Intelligence (WASI). San Antonio, TX: Psychological Corporation.Google Scholar
Wechsler, D (2009) The Wechsler Adult Intelligence Scale—Fourth Edition, 4th ed. San Antonio, TX: Pearson.Google Scholar
Wechsler, D (2011) WASI -II Wechsler Abbreviated Scale of Intelligence, 2nd ed., Vol. 2. San Antonio, TX: Psychological Corporation.Google Scholar
World Health Organization, Mental Health Determinants and Populations Team (2001) Atlas of Mental Health Resources in the World 2001. Geneva: World Health Organization. https://apps.who.int/iris/handle/10665/66910.Google Scholar
Zhang, Y (2021) The role of socioeconomic status and parental investment in adolescent outcomes. Children and Youth Services Review 129, 106186.Google Scholar
Figure 0

Table 1. Prospectively chosen intelligence measures across sites

Figure 1

Table 2. Sociodemographic characteristics of healthy adult participants across sites

Figure 2

Table 3. Mean intelligence scores across sites controlling for biological sex, years of education and SES

Figure 3

Figure 1. Associations between FSIQ scores and sociodemographics. (A) Main effect of WAMI index score (SES) was significant. (B) Full-scale IQ score is positively associated with educational attainment in Brazil (blue), India (maroon), Netherlands (teal) and South Africa (purple) and negatively associated with educational attainment in the United States (green). Main effects of (C) sex were not significant. Figures depicting individual sites in the Supplementary Figure S2. SES, socioeconomic status from the WAMI; FSIQ, full-scale IQ.

Figure 4

Table 4. ANOVA predicting FSIQ scores

Supplementary material: File

DeSerisy et al. supplementary material

DeSerisy et al. supplementary material
Download DeSerisy et al. supplementary material(File)
File 916.1 KB