Response

The paper by Reid et al ^{Reference Reid, Vassilev, Irving, Ojakäär, Jacobson and Lawrence1} highlights that digital assessments show high acceptability and reliability in monitoring cognitive health, thus emphasising the potential for technology integration in dementia care for early intervention and individualised approaches. However, several critical points warrant discussion to enhance the clarity and robustness of the findings.

Rather than using a standard scale for assessing the ease of use and acceptability, such as the system usability scale, ^{Reference Brooke2} authors have used a self-designed Likert questionnaire leading to concerns of validity, sensitivity, reduced credibility and potential bias. Furthermore, the Likert questionnaire has been elaborated as a 10-point Likert scale on the app’s ease of use, asking ‘How would you rate the app in terms of its ease of use? (1 = very difficult; 10 = very easy)’, the length of the assessment, asking ‘How did you find the length of the assessment? (1 = very long; 10 = very short)’ and the difficulty of the assessment, asking ‘How did you find the difficulty of the assessment? (1 = very difficult; 10 = very easy)’. However, the supplementary material with the data analysed for the app’s usability mentions responses ranging from 1 to 5 (1 = disagree/not at all to 5 = agree a lot). Apart from the difference in the scale format and numbering, one version tests ease of use while the other evaluates agreeability, tainting the results obtained.

Studies design their own questionnaires to capture specific knowledge, beliefs, attitudes or behaviours, thereby enhancing the sensitivity of the measurement when compared to an existing instrument. However, there are many issues to consider for developing a questionnaire, such as precision, accuracy, validity, reliability, appropriateness, choosing question and response formats, safeguarding standardisation and ensuring clarity. ^{Reference Boynton and Greenhalgh3} Their versatility allows for applicability across a wide range of constructs, promoting straightforwardness. Moreover, the outcome measured as quantifiable ordinal data aids in the ease of analysis and makes them suitable for use in larger surveys. However, the central tendency bias and acquiescence bias may limit their findings. In addition, it is seen that ordinal data complicates statistical analysis by interpretation variability and forced choices, which may misrepresent true opinions. Cultural differences and individual response styles can further skew results, making it essential to consider these shortcomings for their accurate application. ^{Reference Jebb, Ng and Tay4}

The linear regression for the cognitive tests with age, gender and education shows a different sample size for each test, with 767 participants being analysed for the Snap, Swift and Twist tests, when the initial sample size mentions 756. Moreover, the authors mention that upgradation in difficulty level for the Cast test led to decreased sample size (n = 464) because of exclusion of data collected before upgrading for analysis, but fail to elaborate on the reduced sample size for the Breeze test (n = 458) for which no such upgradation had been performed. The methodology parameters quoted and the result parameters do not seem to correlate in this regard.

Although the use of a constant multiplier in calculation of pillar scores offers benefits such as quantitatively assessing one lifestyle factor in the same manner as others, it does not account for the individual variability in the lifestyle risk factors, for example, individual variability in the effects of sleep requirements and sleep duration on overall well-being, nor does it capture the full spectrum of individual experiences of fatigue with its relation to sleep duration. The lifestyle score did not account for smoking like it did for alcohol, even though smoking status was present in the questionnaire.

While practice effects and their use for assessing cognitive impairment have been highlighted, the availability of the app for practice by the participants during the inter-assessment time intervals has not been highlighted. This in turn would affect the results of practice effects analysed over the course of study. The development of the app for the elderly population required a user friendly interface with availability for use on a laptop/computer, which was a drawback listed in many of the app reviews on the application store.

To conclude, the study by Reid et al ^{Reference Reid, Vassilev, Irving, Ojakäär, Jacobson and Lawrence1} highlights the potential role of digital assessment tools for identifying cognitive changes. The broad inclusion criteria and use of innovative technology to identify dementia risk is quite creditable. However, certain limitations discussed in this review, such as the discrepancies in methodological details (questionnaire format difference, inconsistencies in sample sizes), impact of individual lifestyle factor variability and influence of potential practice effects, highlights the need for more rigorous and transparent analysis. For successful integration of digital tools into dementia care, addressing the above highlighted issues might be beneficial.