Standardization of the Hopkins Verbal Learning Test (HVLT-R) for the Chilean Elderly Population: A Multiple Regression Model Approach

Abstract

Objective

The main goal of this study was to produce linear multiple regression-based normative data of the Hopkins Verbal Learning Test—Revised for the Chilean elderly population.

Method

The study included 229 individuals aged 60–87 years (M = 71.75, SD = 6.64) of both sex (male N = 66, mean age = 72.09 SD = 6.87; female N = 163, mean age = 71.61 SD = 6.57) and educational level (N_{0–12 years} = 68 [29.7%]; N_{13+ years} = 161 [70,3%]).

Results

Age, education, and sex were significantly associated with HVLT-R performance. These demographic variables accounted for 9.4% of the variance in HVLT-R total recall, 11.2% in HVLT-R delayed recall, 8.9% in HVLT-R delayed recognition discrimination index. This study also provides user-friendly percentile rank norms derived from the results of the regression models.

Conclusions

The normative data are presented as regression-based procedures to obtain both direct and derived test scores considering age, education, and sex as predictor variables. This study complies with the guidelines proposed by the Chilean Ministry of Health in its National Plan for Dementias and those of the program for Explicit Health Guarantees by promoting the standardization of instruments that contribute to early diagnosis of neurocognitive disorders in the elderly.

INTRODUCTION

The aging of the population is of global concern. According to the World Health Organization (2020), it is projected that by 2050 the global proportion of older people will double in relation to 2015, rising from 12% to 22%. Likewise, in 2022, the Latin American population over 60 years of age reached 88.6 million, representing 13.4% of the total population, and this percentage is expected to increase to 16.5% by 2030 (Huenchuan, 2018). Furthermore, Chile is among the 3 most aged countries in the region, with 18.1% of older people in 2022 and with an estimated percentage of 32.1% in 2050 (Rojas et al., 2022).

This global change at the demographic level is manifested in different dimensions, one of which is the increase in non-communicable diseases such as mild and major neurocognitive disorders (NCDs; American Psychiatric Association, 2013), whose main non-modifiable risk factor is age (Livingston et al., 2020; Livingston et al., 2024; Tromp et al., 2015).

As a matter of fact, in the Americas region more than 10 million people live with dementia (Comisión Económica para América Latina y el Caribe (CEPAL, 2022) and it is projected that this number will double every 20 years, being Latin America and the Caribbean the most affected regions. In this regard, it is expected that 78 million of older people in these regions will suffer from some type of major NCD in 2030, and that this number will increase to 139 million by 2050 (Prince et al., 2015; World Health Organization, 2021).

Given this context, in 2013, the World Health Organization (WHO) recognizes dementia as a public health priority (World Health Organization, 2013). Later, in 2017, Chile implements the National Plan for Dementias (Abusleme et al., 2017) as a public policy oriented to clinically address NCDs and, at the health care level, the program for Explicit Health Guarantees (GES) approves in 2019 the decree No. 22 named “Alzheimer’s and other dementias” as one of the priority diseases covered by the primary health care system at the national level (Klaassen et al., 2021). The earlier mentioned “National Plan for Dementias” incorporates in its objectives No. 2 and No. 7 the “development of diagnostic systems at different levels of the network with referral and counter-referral protocols” and to consider aging and NCDs as a “priority line of research and innovation” at the national level (Abusleme et al., 2017). The latter makes explicit the need to develop or to validate instruments that allow early diagnosis, as established by the World Health Organization guidelines in relation to the approach to NCDs (World Health Organization, 2021).

Regarding to the aforementioned objective of developing and/or validating diagnostic systems, Chile currently has some standardized screening tests for the detection of NDCs in the elderly, among which we can mention the Montreal Cognitive Assessment (MoCA; Bello-Lepe et al., 2020; Delgado et al., 2017), Mini-Mental State Examination (MMSE; Quiroga et al., 2004) and the Addenbrooke’s Cognitive Examination Revised (ACE-R-Ch; Muñoz-Neira et al., 2012). On the other hand, regarding performance tests for specific cognitive domains, there are normative values available for the Chilean population of the Stroop Color Test (Rivera et al., 2015b), the Trail Making Test (Arango-Lasprilla et al., 2015a), the Rey–Osterrieth Complex Figure Test (Rivera et al., 2015a), verbal fluency tests (Olabarrieta-Landa et al., 2015a), and the Hopkins Verbal Learning Test Revised (Arango-Lasprilla et al., 2015b), among others.

Given that memory complaints are among the earliest and most frequent among the older population with NCDs (Weintraub et al., 2012), the interest of this study focuses on obtaining normative values for performance on the Hopkins Verbal Learning Test - R (Benedict et al., 1998; Brandt, 1991), which is commonly used to assess the mnesic dimension and the anterograde learning abilities. The HVLT-R is a verbal learning test originally created by Brandt (1991), and later revised by Benedict et al. (1998). It has been widely used to assess memory (Jiang et al., 2023; Sáez-Atxukarro et al., 2021; Vicente et al., 2021) and has both excellent psychometric properties and diagnostic efficiency indices (González-Palau et al., 2013; Shapiro et al., 1999). The original HVLT-R study showed an inverse association with age and a direct association with education, even after controlling for age (Benedict et al., 1998). Subsequent studies show that age and education have an impact on both total and delayed recall of the HVLT-R, even in studies conducted with populations from different countries (Hester et al., 2004; Norman et al., 2011; Waldrop-Valverde et al., 2015). In relation to sex, some studies provide evidence of better performance of women in some of the test variables (Ryan et al., 2021; Waldrop-Valverde et al., 2015) which disappears when the comparison is made with samples of people with cognitive deficits (Brunet et al., 2019). The Chilean standardization of the HVLT-R (Arango-Lasprilla et al., 2015b) provides regression-based norms including age, educational level, and sex as predictors, as well as percentile rank norms for total recall and delayed recall stratified by educational level and age groups ranging from 18 to 77 years old and older. However, the mentioned study does not report normative values for the delayed recognition scores, which are of the utter importance to calculate the delayed recognition discrimination index. Additionally, the standards for educational and psychological testing recommends updating test norms with sufficient frequency to permit continued accurate and appropriate score interpretations (American Educational Research Association et al., 2014). In particular, the standard 5.11 states that “test publishers should ensure that up-to-date norms are readily available or provide evidence that older norms are still appropriate” (American Educational Research Association. et al., 2014, p.104). Therefore, the goal of the present study is to provide updated regression-based norms and percentile rank norms for the HVLT-R for both total and delayed recall, and to compensate for the lack of normative values for the delayed recognition scores and, therefore, for the delayed recognition discrimination index in the current available Chilean norms. The reported norms are particularly oriented to assess older populations, and its regression models consider age, years of education, and sex as predictors. Likewise, the percentile rank norms are stratified by educational level and group age, ranging from 60 to 80 years old and older. It is expected that the HVLT-R updated norms will contribute to the objectives of the national plan for dementias and to serve as a useful tool for everyday clinical practice.

METHOD

Participants

The sample was composed of 229 healthy individuals (mean age = 71.75 SD = 6.64) recruited from the FONIS SA20i0095 research project, the senior adult program of the Universidad de Valparaíso, and the community of the region of Valparaíso, Chile. Convenience and snowball sampling methods were used for participant recruitment. Then, participants were stratified by sex, age and educational level. Firstly, we visited elderly community centers and asked university students to contact their relatives that fit the inclusion criteria. We scheduled meetings in which potential participants were informed about the study and clarified their doubts about the study.

Participants had to meet the following inclusion criteria a) to be 60 years old and above, b) regardless of their educational level (i.e., 0–12 years, 13 years or more) to be able to read, write and understand instructions, c) to speak Spanish as their native language, d) and to score 20 or more on the Montreal Cognitive Assessment (Bello-Lepe et al., 2020). Participants diagnosed with neurologic or psychiatric disorders were excluded due to a potential impact on their cognitive performance. Reading, writing and verbal comprehension abilities were evaluated during enrolling to the study, after being assessed by a specialist in neurology. Particularly, participants were asked to write down their full name, date and sign up the informed consent, after terms and conditions of the study were specified. Initially, we were able to contact N = 312 participants of which N = 83 did not met inclusion criteria (i.e., less than 20 points in MoCA N = 54; impaired comprehension and writing abilities N = 17; sensorial impairment N = 3; psychiatric or neurologic disorders N = 9). The final sample was composed as follows: sex (male N = 66, mean age = 72.09 SD = 6.87; female N = 163, mean age = 71.61 SD = 6.57, age group (N_{60–65 years old} = 42 [18%], N_{66–70 years old} = 60 [26%], N_{71–75 years old} = 60 [26%], N_{76–80 years old} = 43 [19%]; N_{81+ years old} = 24 [10%]) and educational level (N_{0–12 years} = 68 [29.7%]; N_{13+ years} = 161 [70,3%]).

According to Bujang et al. (2018), for a 40 items scale, a sample of 134 subjects will suffice to achieve a Cronbach’s alpha reliability index of 0.90 with a statistical power of 0.90 at a significance level of α = 0.05. The study was approved by the scientific ethics committee of the Universidad de Valparaíso (218–20) and followed the Helsinki declaration of ethical principles for research involving human participants (World Medical Association, 2013). Data was collected between March 2022 and June 2023.

Design

We implemented the methodology used by Guàrdia-Olmos et al. (2015) for the development of normative data for ten Spanish-language neuropsychological tests. This normative procedure implements linear multiple regression models and is described in the 2.4. section.

Measures

Hopkins verbal learning test revised version (form 5) (Benedict et al., 1991)

The HVLT-R was created by Brandt (1991) as a brief verbal learning and memory test. It has six alternate forms; therefore, it is useful in populations that require neuropsychological follow-up evaluations (e.g., degenerative diseases) and to evaluate treatment effectiveness (Kreutzer et al., 2011). Each form contains 12 nouns which are distributed into three semantic categories of four words (e.g., professions, sports, and vegetables) to be learned over the course of three learning trials. Approximately 20–25 minutes later, both a delayed recall and a recognition trial are completed. The delayed recall requires free recall of any word of the respective form. The recognition trial is composed of 24 words, of which 12 are target words and 12 are false positives (i.e., 6 semantically related, and 6 semantically unrelated). The total recall score consists of the sum of all three learning trials, while the delayed recall score is equivalent to the number of remembered words after the 20–25 minutes delay (Kreutzer et al., 2011). A retention rate (%) score is calculated by dividing the delayed recall score by the higher number of words remembered in learning trial 2 or 3. Finally, a delayed recognition discrimination index is obtained by subtracting the total number of false positives (i.e., both semantically related and unrelated) from the total number of true positives (Kreutzer et al., 2011).

Normative procedure

In the first place we conducted a Mann–Whitney’s U test by sex for participants’ age as dependent variable. Because age usually affects a participant’s performance in the HVLT-R this analysis was performed to first ensure that there were no significant age differences between men and women that may affect HVLT-R scores. For instance, that women are significantly older than men and that could partially affect HVLT-R performance. Likewise, we conducted a Mann–Whitney’s U test by sex for participants’ years of education as dependent variable, given that years of education may also affect participant’s performance in the HVLT-R. Results show that neither age (U = 5294, p = .575 ns.) nor years of education (U = 4856, p = .887 ns.) showed significant differences by sex (Table 1).

Thereafter, we performed independent t-tests by sex for each HVLT-R score (i.e., total recall, delayed recall, and delayed recognition discrimination index) to determine its incorporation as a predictor into the different multiple regression models. Afterwards, we built the multiple regression models to predict the HVLT-R criterion being normed (i.e., total recall, delayed recall, and delayed recognition discrimination index). Predictors were expressed as a continuous variable for both age and years of education, and as dichotomized variable for sex, when statistical differences were observed (i.e., men coded as 0 and women as 1). Non-significant predictors were excluded from the multiple linear regression models.

Later, we obtained the model coefficients for both the constant and each variable of the predictive model (e.g., age, sex and/or years of education) and through them, the predicted values for each HVLT-R score (i.e., total recall, delayed recall, and delayed recognition discrimination index). Residuals were calculated by subtracting the predicted value from the observed value (e_i = y_i – ŷ_i). Finally, to obtain the percentile for an exact score, we converted the residuals into z scores by dividing the residual value (i.e., e_i) by the standard deviation for the residual obtained in the fitted regression model (z = e_i/SD_e), as described by Van Der Elst et al. (2006, 2012).

Data analysis

Descriptive analyses were conducted to summarize the demographic information of the sample. Later we performed different inferential analyses, consisting in a) Mann–Whitney U tests for detecting sex differences on age and years of education, because normality assumption was not met (Table 1), and also on total recall, delayed recall, and delayed recognition discrimination index (Table 2), b) multiple linear regression analyses and c) conversion of residuals to z scores for elaborating the percentile rank associated to any score stratified by educational level and age-groups.

For the regression analyses, the incorporation of the predictors for each model was made implementing a stepwise procedure (Cohen, 1991). We reported both model coefficients (i.e., non-standardized estimates) and model fit indices (R²). Prior to performing every regression analysis, we tested potential violations of the assumptions for parametric tests under the NHST paradigm (Tinsley & Brown, 2000). We assessed univariate normality (Shapiro Wilk test), collinearity between predictors (Variance Inflation Factor), and autocorrelation (Durbin–Watson test) for each proposed model.

A significance level α = 0.05 was considered for all analyses. Effect sizes for each significant result were also estimated, considering Cohen’s criteria for their interpretation (Cohen, 1988). All analytical procedures were conducted using the statistical Open-Source software Jamovi, version 2.4.11.0 (Jamovi project, 2021).

RESULTS

Descriptive analyses

Table 1 summarizes the demographic information of participants. For categorical variables (i.e., sex) frequency (N) and percentages (%) were provided, whereas for numerical variables (i.e., age, years of education) mean and standard deviations are reported. Additionally, we reported a significance test (i.e., Mann–Whitney U) to evaluate the existence of sex differences in age or years of education that might exert an effect on HVLT-R scores. Results show no statistically significant differences between male and female participant’s age or years of education.

Effect of sex on the HVLT-R total recall, delayed recall, and delayed recognition discrimination index

To analyze the effect of sex and educational level on the HVLT-R total recall, delayed recall, and delayed recognition discrimination index we compared the means of the two levels of each variable (i.e., male vs. female). None of the HVLT-R score distributions met the assumptions for parametric tests, therefore, Mann–Whitney U tests were performed (Shapiro–Wilk’s W p < .001; Levene’s W p < .001). Results are presented in Table 2. All comparisons were statistically significant, except for the delayed recognition discrimination index where no sex differences were observed (p > .05).

Results show that females significantly overperformed males in all HVLT-R scores but delayed recognition discrimination index. The latter allows to incorporate the variable sex as a predictor for total recall and delayed recall in the multiple regression models, coded as a dummy variable with a value of 0 for male, and 1 for female. Effect sizes were equal to r_rb = 0.230 for total recall and r_rb = 0.294 for delayed recall, which should be used as thresholds to be interpreted as medium and large effects in gerontology (Brydges, 2019, p. 3).

Multiple regression models for HVLT-R total recall, delayed recall, and delayed recognition discrimination index

Participants’ age and HVTL-R total recall, delayed recall, and delayed recognition discrimination index was negatively correlated (p < .001). On the contrary, the correlation between years of education and HVTL-R total recall, delayed recall, and delayed recognition discrimination index was positively correlated (p < .001). Because all correlation coefficients were statistically significant, both age and years of education were incorporated as continuous predictors in all three regression models. Assumptions of null hypothesis significance testing were met for total recall (Shapiro–Wilk’s W = 0.993, p = .382; VIF_age = 1.018, VIF_sex = 1.008, VIF_education = 1.022; Durbin–Watson = 1.740), delayed recall (Shapiro–Wilk’s W = 0.970, p = .168; VIF_age = 1.018, VIF_sex = 1.008, VIF_education = 1.022; Durbin–Watson = 1.650), delayed recognition total recall (Shapiro–Wilk’s W = 0.908, p = .572; VIF_age = 1.016, VIF_education = 1.016; Durbin–Watson = 1.938).

The multiple regression model fit for total recall was statistically significant (R² = 0.094, F = 7.80, p < .001). The regression equation to estimate the HVLT-R total recall score is |${\gamma}_{prdicted}=(25.474-( age\times 0.114)+\left( sex\times 2.027\right)+( years of\ education\times 0.298))$|⁠. The model fit for delayed recall was statistically significant (R² = 0.112, F = 9.42, p < .001). The regression equation to estimate the HVLT-R delayed recall score is |$\big({\gamma}_{predicted}=\left(10.321-\left( age\times 0.068\right)+\left( sex\times 1.046\right)\right)+\left( years\kern0.17em of\kern0.17em education\times 0.148\right)\big)$|⁠. Finally, the model fit for the delayed recognition discrimination index was statistically significant (R² = 0.089, F = 10.9, p < .001). However, this model did not include sex as a predictor because no statistical differences were observed for the delayed recognition discrimination index between males and females (U = 4847, p = .115 ns.). The regression equation to estimate the HVLT-R delayed recognition discrimination index score is |${\gamma}_{predicted}=\left(10.490-\left( age\times 0.036\right)+\left( years\kern0.17em of\kern0.17em education\times 0.170\right)\right)$|⁠. Effect sizes ranged from R² = 0.089 to R² = 0.112, which can be interpreted as medium and large according to both Brydges’s (2019) and Cohen’s (1988) criteria. Table 3 summarizes and provides all required elements to easily obtain the standardized values for a subject’s performance in the HVLT-R scales.

Procedure to obtain standardized values for a subject’s performance in total recall, delayed recall, and delayed recognition discrimination index

We explain the procedure of converting raw scores to standardized values (i.e., z scores) through an example. To obtain the standardized value of delayed recall of a woman of 64 years old with 12 years of education and who obtained a score of 10, the predicted value is obtained as follows, (⁠|${\gamma}_{prdicted}=\left(10.321-\left(64\times 0.068\right)+\left(1\times 1.046\right)+\left(12\times 0.148\right)\right)$| = 7.745. As previously mentioned, the predictor sex was coded as a categorical variable with a value of 0 for male, and 1 for female. Later, to obtain the residual value of the equation, we subtract the predicted value from the obtained score, which in this case is, |${e}_i=\left(8-7.745\right)=0.225$|⁠. Finally, to convert this result into a z score the obtained residual is divided by the residual’s standard deviation (i.e., SDe). In this case, for delayed recall the SDe value provided in Table 3 is 2.1. Thus, the z score is calculated as, |$\left({z}_i=\left(\frac{0.225}{2.1}\right)=0.12\right)$|⁠. Results are expressed in standard deviation units, then, the performance of a 72-year-old woman with 13 years of education is 0.12 standard deviation units above the mean. This value is equivalent to a percentile rank of 54.

For those clinicians who are not familiarized with statistics models, table 4 provides a more user-friendly conversion method with percentile rank norms for HVTL-R total recall, delayed recall, and delayed recognition discrimination index. However, we strongly encourage to use the previously described conversion procedure which provides more precise estimates than those provided in table 4, because incorporates the exact weight (i.e., regression coefficients) of all predictors (i.e., age, sex, years of education) to convert the raw scores into standardize values.

DISCUSSION

The purpose of this study was to obtain normative data of the HVLT-R for the Chilean population of people aged 60 years or older, stratified by age, sex, and education, using a methodological approach based on multiple linear regression models (Arango-Lasprilla et al., 2015b; Guàrdia-Olmos et al., 2015; Van Der Elst et al., 2006, 2011). This approach allows for a more accurate assessment adjusted to the specific characteristics of each subject, allowing for a more appropriate and fairer evaluation based on the sociodemographic characteristics of the participants. The final models explained 9.4% of the variance of the total recall scores, 11.2% of the variance of the delayed recall, and 8.9% of the variance of the discrimination index of the delayed recognition task. The observed results for total recall and delayed recall are inversely correlated to age and directly correlated to education, which is consistent with the previous validation of the HVLT-R for the Chilean population conducted by Arango-Lasprilla et al. (2015b). Our findings are also consistent with previous studies that emphasize the influence of sociodemographic variables on the cognitive performance in older people (Friedman et al., 2002; González-Palau et al., 2013; Hester et al., 2004; Jiang et al., 2023; Ryan et al., 2021). Indeed, age, education, and sex have shown to be significant predictors of both total and delayed recall performance in the HVLT-R (Benedict et al., 1998; Vanderploeg et al., 2000). In this regard, Vanderploeg et al. (2000) and Benedict et al. (1998) reported that both age and sex had an influence on total and delayed recall performance. Our results reveal that age, education and sex accurately predict both total and delayed recall scores of the HVLT-R, but not the delayed recognition discrimination index. The latter suggests that both males and females’ performance on discriminating and recognizing learned verbal material is almost equal. A potential explanation for this phenomenon is that the presence of the stimuli (i.e., words) in the delayed recognition task facilitates the activation of the retrieval processes linked to the dorsolateral prefrontal cortex (Tirapu-Ustárroz & Muñoz-Céspedes, 2005; Tulving & Markowitsch, 1997).

A major contribution of our study with respect to that of Arango-Lasprilla et al. (2015b) is that it provides normative values for the delayed recognition discrimination index, which is clinically relevant to differentiate memory impairment associated with deficits in memory retrieval versus consolidation processes, associated with prefrontal and hippocampal lesions, respectively (Tirapu-Ustárroz & Muñoz-Céspedes, 2005; Tulving & Markowitsch, 1997). Additionally, in Arango-Lasprilla et al. (2015b) education was included as a dummy coded categorical predictor (i.e., 0 = 01–12 and 1 = 13 or more years of education), whereas the present study considers education as a numerical continuous predictor and, therefore, allows for more accurate estimations of total, delayed and delayed recognition scores. As the example provided earlier, the performance of a 72-year-old woman with 13 years of education is 0.12 standard deviation units above the mean. This value is equivalent to a percentile rank of 54. However, the performance of the same person using Arango-Lasprilla’s (2015b) norms would be estimated in 0.73 standard deviation units above the mean. This value is equivalent to a percentile rank of 76. These differences can be, mainly attributed to differences in the predictors used in both studies, to the fact that the age range is narrow in our study and that, as mentioned, our study considers education as a continuous variable instead of a dummy-coded variable.

Considering that age is the main non-modifiable risk factor for the onset of NCD (Livingston et al., 2020; Livingston et al., 2024; World Health Association, 2012) our study focused on obtaining normative HVLT-R values for the population of people aged 60 years and older. This methodological decision becomes even more relevant if we assume that the mnesic dimension is one of the first to be affected in both mild and major NCD, with a prevalence of about 70% of cases (OMS & OPS, 2013).

Another aspect that gives relevance to our study is the substantial lack of norms for Spanish-speaking population in Latin America (Guàrdia-Olmos et al., 2015), which has motivated various researchers to validate assessment tests for other cognitive domains (see Arango-Lasprilla et al., 2015a; del Cacho-Tena et al., 2023; Guàrdia-Olmos et al., 2015; Olabarrieta-Landa et al., 2015a; Olabarrieta-Landa et al., 2015b; Rivera et al., 2019; Rivera et al., 2015a; Rivera et al., 2015b; Vicente et al., 2021).

An important implication of the present study is to provide up-to-date norms for the interpretation of the Hopkins Verbal Learning Test R, which according to the standards for educational and psychological testing should be regularly revised to ensure accurate score interpretations (American Educational Research Association et al., 2014). Another implication of our study is that implements regression-based norms, which contributes to address some issues related with conventional norming by, for example, reducing sampling errors, eliminating age-related biases, filling gaps in norm tables, and requiring smaller, more cost-effective sample sizes (Lenhard et al., 2019). This approach to the validation of neuropsychological tools is becoming increasingly used in the field of neuropsychology (Eliassen et al., 2023; Karstens et al., 2024; Lorentzen et al., 2021; Scheffels et al., 2023; Vicente et al., 2021).

Among the limitations of this study is the lack of representation from all regions of Chile. Previous validation studies (Bello-Lepe et al., 2020) recruited participants from the north, center and south areas of our country, which contributes to improve the interpretability of the predicted scores based on factors related to multiculturality and other sociodemographic variables.

Another important limitation of the present study was the overrepresentation of participants with high educational level, which may overestimate the average years of education of the overall population. However, there are some significant demographic changes in the Chilean population of 60+ years old that may partially explain the sample’s elevated educational level. According to a recent publication of the center of studies on aging and elderliness (Observatorio del Envejecimiento, 2022, p.2) the proportion of older persons who have attained higher professional or technical education has almost doubled between 1990 and 2020. The proportion of 60+ years old persons with high education levels increased from 26,9% to 54,7%. Likewise, in the same time period, it also occurred a large decrease in those older persons who only reached basic education or did not attain formal education at all. This decrement was from 73% in 1990 to 45,2% in 2020. Furthermore, at a regional level (i.e., Valparaíso region), we observed a similar phenomenon, and the proportion of 60+ persons with at least 12 years of education is 49%, with an increment of 108,6% between 2018 and 2020, ranking second nationwide in terms of years of education for the 60+ years old segment. Therefore, these changes may have been partially reflected in the average years of education observed in our actual sample.

Nevertheless, in order to mitigate or, at least, minimize a potential underestimation of those persons with low educational levels (i.e., 12 years or less), we proposed a procedure to obtain more accurate predictions. First, we calculated the standard deviation of the residuals attributed to the years of education. These values were estimated to be SDe = 4,4 for total recall, SDe = 2,2 for delayed recall, and SDe = 1,6 for the delayed recognition discrimination index. Hence, for a 60+ years old person with 12 or less years of education we recommend adding 4 extra points to the obtained total recall score, and to add 2 extra points to both, the obtained delayed recall score and the delayed recognition discrimination index score. Afterwards, the corrected scores should be incorporated into the regression equations. For those persons with high educational levels no scores corrections should be made. Finally, when using the percentile rank norms, the corrected scores should be used only if the person’s educational level is equal or less than 12 years. As an example, obtained scores of total recall = 24, delayed recall = 8, and delayed recognition discrimination index = 7, will be converted to the corrected scores of 28, 10 and 9 points for those persons with 12 or less educational years. The described procedure will contribute to reduce any potential underestimation of 60+ persons’ performance with low educational levels. It should be declared that similar correction procedures have been implemented in other neuropsychological tests like the Montreal cognitive assessment (Nasreddine et al., 2005), in which one point is added to the total score in the case of persons with less than 12 years of education.

At last, but not least, increasing the sample size would also contribute to minimize the random error associated to the estimation of the regression coefficients and, therefore, allowing for even more accurate score predictions. Further studies should improve these relevant aspects of validation studies.

Finally, this study complies with the guidelines proposed by the Chilean Ministry of Health in its National Plan for Dementias (Abusleme et al., 2017; Gajardo & Abusleme, 2016) and the program for Explicit Health Guarantees (Ministerio de Salud & Subsecretaría de Salud Pública, 2019) by promoting the creation and/or standardization of instruments that will contribute to early diagnosis in the field of NCD in older people. The latter will favor the implementation of both pharmacological and non-pharmacological therapeutic interventions, focused on the patient and his or her social environment. Forthcoming studies will incorporate measures in other cognitive domains, such as attention, language, executive functions, among others.

FUNDING

This work was supported by Agencia Nacional de Investigación y Desarrollo, Ministerio de Salud, Gobierno de Chile (grant number FONIS SA20i0095).

CONFLICT OF INTEREST

None declared.

In loving memory of Sebastián Bello-Lepe, PhD, March 1 1988–December 25 2023

References