Mini-Cog versus Codex (cognitive disorders examination) Is there a difference?

Abstract. Mini-Cog and Codex (cognitive disorders examination) are brief cognitive screening tests incorporating word-recall and clock drawing tests. Objective: To assess and compare the screening accuracy of Mini-Cog and Codex for diagnosis of dementia and mild cognitive impairment (MCI) in patients attending a dedicated cognitive disorders clinic. Methods: Tests were administered to a consecutive cohort of 162 patients, whose reference standard diagnoses based on clinical diagnostic criteria were dementia (44), MCI (26), and subjective memory complaint (92). Results: Both Mini-Cog and Codex had high sensitivity (>0.8) for dementia diagnosis, but Codex was more specific. For diagnosis of MCI, Mini-Cog had better sensitivity than Codex. Weighted comparisons of Mini-Cog and Codex showed only marginal net benefit for Mini-Cog for dementia diagnosis but larger net benefit for MCI diagnosis. Conclusion: In this pragmatic study both Mini-Cog and Codex were accurate brief screening tests for dementia but Mini-Cog was better for identification of MCI.

RESUMO. Mini-Cog e Codex são testes breves de triagem cognitiva incorporando testes de recuperação de palavras e desenho de relógio. Objetivo: Avaliar e comparar a precisão da triagem de Mini-Cog e Codex (exame de distúrbios cognitivos) para o diagnóstico de demência e comprometimento cognitivo leve (CCL) em pacientes atendidos em uma clínica dedicada a distúrbios cognitivos. Métodos: Os testes foram administrados a uma coorte consecutiva de 162 pacientes, cujos diagnósticos padrão de referência com base em critérios clínicos de diagnóstico foram demência (44), CCL (26)  Mini-Cog consists of a three word recall task and a clock drawing task. 3 In the standard scoring system, a score of zero or Larner Mini-Cog vs. Codex three on the word recall task leads to categorization as "dementia" or "no dementia"; for the intermediate scores on word recall, 1 or 2, performance on the clock drawing task is then taken into account: if normal or abnormal the patient is categorized as "no dementia" or "dementia", respectively. The cognitive disorders examination or Codex is a two-step decision tree for diagnostic prediction which incorporates the three-word recall and spatial orientation components from the MMSE along with a simpli-fied clock drawing test (CDT). 4,5 The endpoint values of the four terminal nodes have different probabilities of dementia diagnosis: categories A-D, respectively with very low, low, high, and very high probability of dementia ( Figure 1A).
There is item overlap between Mini-Cog and Codex. Rewriting Mini-Cog as a decision tree ( Figure 1B) and comparing this with Codex, it is evident that the two sequential steps of Mini-Cog are applied simultaneously in Step 1 of Codex ( Figure 1). Do Mini-Cog and Codex perform similarly in clinical practice? The primary aim of this study was to compare the accuracy of these two CSIs in screening for dementia, a comparison which has not, to the best of the author's knowledge, been reported hitherto in the literature. The secondary, exploratory, aim was to compare accuracy in screening for mild cognitive impairment (MCI), an important clinical distinction, although it is acknowledged that neither of these CSIs was specifically designed for the identification of MCI.

METHODS Participants
Data from a consecutive patient cohort referred to a dedicated cognitive function clinic based in a secondary care setting (neurology clinic) over a fixed time period of nine months (February-November 2012 inclusive) 6,7 and who were administered MMSE and CDT were re-analysed. Criterion diagnoses were dementia, mild cognitive impairment (MCI), or subjective memory complaint (SMC) by judgment of an experienced clinician, the former two diagnoses based on standard diagnostic criteria for dementia (DSM-IV; study data collection preceded publication of DSM-5) and MCI. 8 Subjects gave informed consent and the study protocol was approved by the institute's committee on human research.

Procedures
For Mini-Cog, categorical data (dementia/no dementia) were derived using the standard scoring method ( Figure  1B). 3 For Codex, categorical data were derived from the decision tree ( Figure 1A) with differing probabilities of diagnosis (A = very low, B = low, C = high, D = very high); categories C and D were taken to be indicators of cognitive impairment, as per the index publication. 4 Neither Mini-Cog nor Codex categories were used in reference diagnosis to avoid review bias.

Statistical analysis
STARDdem guidelines for reporting diagnostic test accuracy studies in dementia were observed. 9 Standard summary measures of discrimination were calculated: sensitivity and specificity, Youden index (Y), positive and negative predictive values (PPV, NPV), predictive summary index (PSI), correct classification accuracy (Acc), net reclassification improvement (NRI), positive and negative likelihood ratios (LR+, LR-; classified according to Jaeschke et al.), 10 diagnostic odds ratio (DOR), and clinical utility indexes (CUI+, CUI-; classi-fied according to Mitchell). 11 In addition, a number of recently described unitary metrics 12 were calculated: the "likelihood to be diagnosed or misdiagnosed" which is the ratio of the number needed to misdiagnose (NNM = 1/(1 -Acc)) to either the number needed to diagnose (NND = 1/Y) or the number needed to predict (NNP = 1/PSI), where LDM >1 is desirable; 6,12 and the summary utility index (SUI = CUI+ + CUI-) and the number needed for screening utility (NNSU = 1/SUI), with values classified as previously reported. 12,13 Mini-Cog and Codex were compared by means of weighted comparison performed as per the method of Moons et al.: 14 where π = prevalence; FP = false positives; and TP = true positives. Δ sens and Δ spec are the differences in the sensitivity and specificity of the two tests at the chosen test cut-offs. The relative misclassification cost (FP/TP), a parameter which seeks to define how many false positives a true positive is worth, was set at FP/TP = 0.1, following previous usage, 15

RESULTS
Baseline demographic data from the study are shown in Table 1, along with the distribution of observed Mini-Cog and Codex categories. Final diagnoses in the dementia group were Alzheimer's disease or mixed Alzheimer's disease and cerebrovascular disease (31), frontotemporal dementia (6), dementia with Lewy bodies (4), progressive supranuclear palsy (2) and alcohol-related dementia (1). In the MCI group, final diagnoses were amnestic MCI (19; single domain 8, multiple domain 11) and non-amnestic MCI (7; single domain 1, multiple domain 6).
Patient numbers with positive or negative Mini-Cog and Codex tests as a function of the reference diagnosis are shown in Table 2.
Examining the primary study aim, for the diagnosis of dementia versus no dementia the measures of discrimination were similar for Mini-Cog and Codex, with Mini-Cog slightly better for sensitivity, NPV, and LR-, but Codex was slightly better on all other measures, including the unitary measures (Table 3, left hand columns).   Weighted comparison showed a minuscule net benefit for Mini-Cog versus Codex for the primary study aim of dementia diagnosis, with less than 6 additional TP patients identified per 1000 screened (Table 4).
Weighted comparison for the secondary study aim of MCI diagnosis showed a larger net benefit for Mini-Cog versus Codex, resulting in an equivalent increase of around 56 extra MCI cases detected per 1000 tested (Table 4).

DISCUSSION
Both Mini-Cog and the Codex decision tree are quick and easy to use. The study data showed that for the primary study aim of dementia diagnosis both Codex and Mini-Cog had good and similar metrics.
For the secondary study aim of MCI diagnosis, both tests were poor at differentiating MCI from SMC, an unsurprising finding since neither test was designed for this purpose, and as previously noted for Codex in this dataset 6,7 and in independent studies of Codex. 4,5,16,17 Mini-Cog appeared better than Codex for MCI diagnosis. Reasons for this disparity are evident from perusal of data in Table 1: Mini-Cog classified more patients as impaired (category "dementia" = 43.2%) than Codex (categories C and D = 35.2%) suggesting it has greater sensitivity for cognitive impairment than Codex, which evidently classifies some impaired patients as having a low probability of impairment (false negatives).
A previous attempt to circumvent the lack of sensitivity of Codex for cognitive impairment short of dementia by means of simple modifications of the decision tree proved unsuccessful. 7 The application of a simple logical "And" rule in the first Codex step (also known as conjunctive combination, or "believe the negative") might be anticipated to reduce sensitivity, NPV, and LR-, as observed with other applications of the "And rule". Application of the "Or" rule (compensatory combination; "believe the positive") may be advantageous for case finding as this approach generally improves sensitivity. 18 Limitations of this study include use of clinical diagnostic criteria for dementia and MCI and the crosssectional design which risks incorrect categorization of cases. Use of clinico-biological criteria incorporating imaging and CSF biomarkers (not available to this clinic) and longitudinal follow-up for delayed verification of diagnosis might circumvent these problems. The spectrum of final diagnoses reflects the selection bias encountered in a clinic based in a neurology centre, related to the relatively young age of the patient population (median 61 years), hence possibly limiting the generalizability of the findings; the diagnostic spectrum is likely to differ in studies undertaken in other secondary care settings such as old age psychiatry or geriatric medicine.
Deriving Mini-Cog scores retrospectively from lengthier tests has the potential to introduce bias, although this approach has been noted in previous Mini-Cog studies included in systematic reviews. 19 This methodological limitation may limit the generalizability of findings since CSIs are not typically used in this way in clinical practice. In addition, the sample size in this study was relatively small, and no power calculation to  20 When clinical assessments are time limited, Mini-Cog may be a possible option if the clinician wants to avoid false negative diagnoses of cognitive impairment, i.e. requires a high sensitivity test. Another very short screening test which, like Mini-Cog and Codex, incorporates recall and clock drawing is the Rapid Cognitive Screen, 21 which may also be combined with simple cate-gorical clinical signs which may indicate presence of cognitive impairment (Triple Test). 22 However, the potential advantages of test brevity should be weighed against the evidence suggesting that CSI length (number of test items) correlates positively with measures of diagnostic accuracy. 23 Other CSIs with slightly longer application times, around 10 minutes, may also be effective for the identification of MCI, particularly the Montreal Cognitive Assessment and the Mini-Addenbrooke's Cognitive Examination 24,25 and the Qmci screen. 26