Psychometric characteristics of the full and short versions of the IQCODE-BR among low income elderly persons with a low educational level

Objective: To study the psychometric characteristics of the Brazilian Portuguese version of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE-BR) using a sample of low income elderly persons with a low educational level, and compare the full and short versions of the questionnaire. Method: A cross-sectional study of a convenience sample of a population with a low educational level was performed. The IQCODE-BR was applied to the informants of 87 elderly persons (60-90 years old), who were triaged by psychiatrists and neurologists for the diagnosis of depression, mild cognitive impairment (MCI), and dementia. Results: The median age of the sample was 72 and the majority were women (72.4%). A total of 31 (35.6%) were illiterate, 30 (34.5%) Palavras-chave:


INTRODUCTION
[3] Dementia is one of the most significant causes of morbidity and mortality among the elderly.It is a chronic syndrome whose main characteristics are the decline of memory and other intellectual functions, such as language, praxis, the ability to recognize and identify objects, abstraction, organization, planning and sequencing, as well as changes in behavior or personality, impairing the psychosocial development of the individual. 4e vast majority of cases of dementia cases are degenerative, and the syndrome is the main cause of Alzheimer's disease.There is still no satisfactory treatment.The few drugs that can slightly modify the progression of the disease are more effective in the early stages. 5Mild Cognitive Impairment (MCI) is manifested by a change in cognition compared to the previous level of the patient, with impaired performance in one or more of the cognitive domains that are expected for the individual's age and education, but with the retention of independence in functional skills. 6,7 this context, validation tools that can identify the disease in its early stages are extremely important.0][11] It consists of 26 questions and is applied to a family member or caregiver who has had contact with an elderly individual for over ten years.This instrument was translated and validated for Brazilian use (IQCODE-BR) 12 and a reduced version was created, composed of 16 questions taken from the full version.
Studies have indicated the importance of informant reporting as a fundamental part of the criteria for the diagnosis of dementia. 13,14It is essential to evaluate the history of cognitive decline through the observations of a relative, especially when assessing an individual with a low education level, whose performance in neuropsychological tests may be compromised.Functional tests should be combined with cognitive assessment to improve diagnostic accuracy. 15e aim of the present study was to evaluate the psychometric characteristics of IQCODE-BR in a sample of low income elderly individuals with a low educational level, comparing the full and reduced versions.

METHOD
This study is a part of the Programa de Envelhecimento Cerebral ("Brain Aging Program") (PENCE) of the Estratégia Saúde da Família ("Family Health Strategy") (ESF) of Porto Alegre in collaboration with the Instituto de Geriatria e Gerontologia ("Institute of Geriatrics and Gerontology") of the Pontifícia Universidade Católica do Rio Grande do Sul (IGG-PUCRS).PENCE was developed from an EMI-SUS crosssectional study of the population registered with the Porto Alegre ESF. 16The participants of the present study are part of this sample, which is characterized by low income individuals.Of these, 7.0% have no income, 54.7% earn less than the minimum wage, 25.0% earn the equivalent of one to two monthly minimum wages, 5.0% earn the equivalent of two to four times the monthly minimum wage and approximately 2.0% earn the equivalent of four to six monthly minimum wages, as described in other studies using the same population group. 17,18total of 87 elderly individuals aged 60 to 90 years treated at the Ambulatório de Envelhecimento Cerebral ("Brain Aging Clinic") (AMBEC) in the period from March to December 2013, who were accompanied by an informant they had known for at least ten years, were included in the study.The AMBEC was created specifically at the Neurology Service of the Hospital São Lucas of PUCRS to care for the mental health of elderly individuals registered with PENCE.
Patients diagnosed with central nervous system diseases other than MCI or dementia or with psychiatric axis I disorders other than depressive disorder or anxiety were excluded from the study.
All patients underwent at least one neurological consultation and one psychiatric consultation.In the neurological consultation, the patients were cognitively evaluated using Addenbrooke's Cognitive Examination Revised (ACE-R).Where required, the patients were referred for neuropsychological assessment using the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) battery of evaluations.In psychiatric consultation, patients were evaluated with the reduced 15 item Portuguese version of the Geriatric Depression Scale (GDS) and the Portuguese mini-version 5.0 of the Mini-International Neuropsychiatric Interview (MINI), a screening tool for Axis I psychiatric disorders.
The diagnosis of MCI and dementia followed the recommendations of the National Institute on Aging -Alzheimer's Association (NIA-AA) published in 2011. 19Diagnoses of more severe depression were made in consultation with a psychiatrist, based on the DSM-IV-TR criteria. 20oth the psychiatrist and the neurologist had no knowledge of the IQCODE-BR result.The nurse who applied the IQCODE-BR in relation to neurological and psychiatric diagnosis was also unaware of the result.
The IQCODE-BR is a questionnaire for the detection of cognitive decline based on the reports of informants.The full version consists of 26 questions, of which 16 are included in the reduced version.The questions are answered based on the Likert scale, with five options (1 much improved; 2 a bit improved; 3 not much change; 4 a bit worse, 5 much worse).The end result is obtained by the weighted sum of the items, which is then divided by the total range of items.The final score of the two versions ranges from 1 to 5. The questionnaire was applied by the same researcher and took approximately 15 minutes.The full version was applied to all participants, and analysis of the full and reduced versions was performed at a later date.
The data was entered into a database specially developed for the project using the File Maker Pro 11 program.Statistical analysis was performed using the SPSS 17 program.Frequencies, means and standard deviations were used for descriptive statistics.Pearson's chi-squared test was used to compare the categorical variables (demographic data) of the groups.When more than 25% of households had an expected value below 5, Fisher's exact test was used.To compare the means between groups, variance analysis (one-way ANOVA) with the Bonferroni post-hoc test was used.To control the confounding effect, the comparison of the IQCODE-BR mean values between the groups was adjusted for age and education through multiple linear regression probability (logit).The Pearson correlation coefficient was used to analyze the association between the IQCODE-BR and the MMSE and ACE-R values, and the Spearman correlation coefficient was used to analyze the association between the IQCODE-BR and GDS values.Exploratory factor analysis was used for analysis of construct validity, with the application of varimax rotation, and internal consistency was assessed with Cronbach's alpha.To study the accuracy of IQCODE-BR and the best cut-off points for the diagnosis of MCI and dementia, ROC curves were constructed and sensitivity and specificity were calculated.
Accompanying persons/informants who participated in the study signed a Free and Informed Consent Form and responded to the full version of IQCODE-US, with 26 questions.

RESULTS
The mean age of the 87 elderly respondents was 72.1 (SD = 7.5) years.The data presented in Table 1 is related to sociodemographic characteristics.It is noteworthy that most of the elderly individuals were female (72.4%) and that 35.6% were illiterate.The elderly were divided into four groups: 30 with dementia, 20 with MCI, 21 with depression and 16 with no such diagnoses.It was observed that there was an association between MCI and dementia and advanced age and low education.There was no significant difference in the results based on gender.The distribution of the final results of the full and reduced versions of IQCODE-BR for each diagnostic group is shown in Table 2. Full and reduced mean IQCODE-BR results were, respectively, 3.18 and 3.15 for elderly persons without cognitive impairment and depression; 3.38 and 3.39 for elderly persons with depression; 3.48 and 3.48 for elderly persons with MCI; and 3.78 and 3.83 for elderly persons diagnosed with dementia.A statistically significant difference was observed between the groups (p <0.001).Post hoc analysis revealed there was no significant difference between the groups with depression and MCI, but found that these groups differed significantly from normal individuals and elderly individuals with dementia.
The full and reduced versions of the IQCODE-BR had Cronbach's alpha values of 0.906 and 0.908 respectively.Individually, all the questions had values greater than 0.90, except for question 25 of the reduced version, which had a Cronbach's alpha value of 0.898.Exploratory factor analysis resulted in an overall variance percentage of 57.8 for the full version and 65.3 for the reduced version, with the identification of five and four possible domains, respectively.When analyzing the questions to which each domain was linked with most affinity, it was observed that many questions belong simultaneously to more than one domain.There are not, therefore, clearly defined individual domains.
The correlation between IQCODE-BR (complete and reduced version) and the instruments assessing cognition, ACE-R and MMSE, was moderate and statistically significant, with a correlation coefficient of -0.492 and -0.483 for ACE-R and -0.527 and -0.512 for MMSE respectively.

DISCUSSION
Several studies have shown that age and educational level act as risk factors for cognitive decline and the development of dementia, and that a higher educational level acts as a protective factor against the same. 21,22The study population, despite being a convenience sample, included elderly people with and without mental health problems registered with the Porto Alegre ESF.In the present study there was an association between age and the educational level of the elderly persons studied.The frequency of older patients with lower educational levels was higher among patients with MCI and dementia.
There was a strong association between the IQCODE-BR results and the diagnosis of dementia, given that similar results were obtained for MCI and depression.There was no significant difference between the means of the MCI and the depression groups, which indicates that depression can interfere with the tracking of dementias.Studies have shown that depression is related to the detection of MCI 23,24 and is often a confounding factor for its diagnosis.
The internal consistency of the present study, both when using the full and the reduced version of the questionnaire, was also strong, with a score of greater than 0.90 for the sample surveyed here, and in wider literature. 25,26th respect to exploratory factor analysis, it was not possible to group the questions into appropriate separate domains, as most fell into more than one domain.As such, exploratory factor analysis could not be fully completed as the questions cannot be classified into the correct domains, suggesting that the affinity between the questions is directly interconnected.This finding was even more significant in the reduced version.
No IQCODE-BR domains were identified as the questionnaire was not based on such divisions, and is an instrument for investigating activities of daily living compared with cognitive performance over the previous ten years.Most issues correlate with more than one domain.It is inferred that because of their direct affinity, there is no way to separate them into domains.
The correlation coefficient with the MMSE and ACER was moderate, with values below r = -0.630.[29] The full and reduced versions had similar psychometric characteristics, meaning the reduced version can be used without a significant loss of quality of the instrument among this population, although other instruments should be used for diagnostic purposes. 30 assessing accuracy, a larger ROC curve area was obtained when the elderly diagnosed with depression were replaced with those with MCI.Although different cut-off points were established for the full and reduced versions, both versions proved effective at tracking MCI and dementia.The present study suggests that the IQCODE-BR is able to detect dementia in individuals with a low educational level at an early stage, but that its findings can be influenced by a diagnosis of depression.Therefore, for the reduced version cut-off points of 3.22 to identify CCL (when depressive symptoms are excluded) and 3.48 to identify dementia (regardless of the presence of depression) are suggested.
][33][34] The aim of the present study was to study data from a population with a low educational level.
Because it is a convenience sample, with data from an ESF outpatient referral clinic, the results of this study refer to a low income population with a low educational level.The findings of the present study show that this easy to apply, accurate instrument can be used as a diagnostic tool to detect dementia at an early stage by ESF nurses and technicians in a primary care environment.

Figure 1 .
Figure 1.ROC curve showing relationship between sensitivity and specificity of scores of complete and reduced versions of IQCODE-BR for the diagnosis of mild cognitive impairment (MCI) and dementia.Porto Alegre, RS, 2013.

Figure 1
Figure 1 shows the ROC curves generated for the diagnosis of MCI and dementia among all the elderly persons (A and B), and excluding elderly persons with depression (C and D).A cutoff point of 3.22 was identified for diagnosis of MCI with the reduced version when patients with The researchers followed the recommendations of Resolution nº 466/2012 of the National Health Council of the Ministry of Health.The project was approved by the Ethics Research Committee of the Pontifícia Universidade Católica do Rio Grande do Sul, registered under n° 228.937, dated 13 March 2013.

Table 2 .
Mean, standard deviation, median and ranges of final results of complete and reduced versions of IQCODE-BR, in accordance with diagnosis.Porto Alegre, RS, 2013.
CCL= mild cognitive impairment; m= mean; sd= standard-deviation; med= median; *p value calculated by variance analysis (one way ANOVA); % value of p corrected for age and educational level.