A comparison of electronic and paper versions of the Quick Mild Cognitive Impairment (Qmci-CN) screen

Abstract

Background

There is a growing focus on digitalizing screening instruments for use on computerized, mobile devices. However, few studies have compared the performance of electronic and traditional ‘pen and paper’ formats.

Objective

This study aimed to compare the performance of electronic and traditional ‘pen and paper’ formats of the Chinese version of Quick mild cognitive impairment screen (Qmci-CN) screen in adults.

Methods

A convenience sample of 86 adults aged 18 years were included. Participants were randomly assigned to two groups that underwent the eQmci-CN and pQmci-CN screens in counterbalanced order. Patients completed a questionnaire detailing their experiences and preferences. Total and subscale scores and administration times were recorded.

Results

A strong, statistically significant correlation (r = 0.77) was found between eQmci-CN and pQmci-CN scores for all participants. Analysis comparing the first administration only (eQmci-CN versus pQmci-CN) found no statistically significant difference in total scores, except for one subtest registration (p = 0.023), where participants scored lower on the eQmci-CN. Administration times were similar, though the pQmci-CN was significantly shorter than eQmci-CN (287.41 s versus 302.78 s, respectively, p = 0.005).

Conclusions

The eQmci-CN demonstrated strong correlation with the pQmci-CN in this sample of Chinese adults. However, further research is required to examine these findings in a broader population to enhance the reliability and external validity of the results.

Introduction

As society ages, cognitive impairment, including both mild cognitive impairment (MCI) and dementia, has become an increasingly recognized as a global public health issue. ¹ By 2050, the total number of people with dementia is projected to reach 152 million, more than triple the number reported in 2015. ² MCI refers to a condition characterized by cognitive decline that does not meet the diagnostic criteria for dementia, with preserved independence in activities of daily living (ADLs).^3,4 MCI is regarded as a prodrome to dementia and its prevalence, like that of dementia increases with age.^5–7 Early detection of cognitive impairment, particularly at the MCI stage, is crucial for initiating lifestyle interventions and, in future, potential disease modifying treatments.^8–16 Therefore, there is a pressing need for simple, reliable, accurate and accessible cognitive screening tools to identify individuals who require further assessment.

Currently, several cognitive screening instruments are available. ⁵ However, these tools have some limitations, including variations in administration time, sensitivity and specificity, which restrict their widespread use.^16,17 According to a recent consensus study by Morley et al., an ideal screening instrument should take less than 5 min to administer. ¹⁶ Such a tool should be brief, highly accurate, sensitive and specific, easy to administer with minimal training, and free from copyright charges.^16,17

In recent years, computerized cognitive screenings instruments have gained prominence due to their ability to provide quantitative, high-precision measurements that surpass traditional pen-and-paper tests. ¹⁸ Widely used tools such as the Montreal Cognitive Assessment (MoCA), have been developed into electronic formats and have demonstrated reliability and accuracy. ¹⁹ Compared to traditional pen and paper testing, computerized assessments offer standardized administration, reduced inter-rater variability, and objective data outputs.^18,20–22

The Quick Mild Cognitive Impairment (Qmci) screen is a brief and accurate cognitive screening instrument designed to differentiate MCI from normal cognition and early dementia. ²³ The Qmci screen has six subtests: Orientation, registration, a clock drawing test, a delayed recall section, a test of verbal fluency and a test of logical memory (immediate verbal recall of a short story). It takes between 3–5 min to complete and is shorter than many other cognitive screening instruments.^24,25 Despite this, it has excellent accuracy and is highly sensitive and specific at detecting MCI. ¹⁷ Therefore, the Qmci screen may be particularly useful in busy clinical settings such as in primary care. ²⁶ The Chinese version of the Qmci screen (Qmci-CN), which was translated and adapted by Xu et al. in 2020, has an optimal cut-off score of 55 out of 100 for cognitive impairment. ²⁵ In 2021, Xu et al. found that the Qmci-CN is accurate in identifying post-stroke dementia (PSD) and separating PSD from post-stroke cognitive impairment no dementia (PSCIND) in patients post-stroke during rehabilitation and is comparable to MoCA-CN, albeit with a significantly shorter administration time. ²⁷

The electronic Qmci-CN (eQmci-CN) is based on the original paper-based test, without any changes in the content or scoring methods. It runs on desk-top, tablet and smart phone devices. The interface incorporates a guide with questions and answers. The eQmci-CN features pre-recorded verbal instructions accompanied by on-screen instructions and practice questions using a stylus to support self-assessment by patients.

Given the healthcare sector's shift toward digital precision medicine, this study evaluates the eQmci-CN's design, performance, and reliability in Chinese populations, with a focus on its quantitative advantages over conventional methods.

Methods

Participants

A total of 95 participants were recruited to complete both the electronic (eQmci-CN), and pen and paper (pQmci-CN) versions of Qmci-CN between January 3, 2021 and the May 23, 2021, in Guangzhou, China. Participants were consecutive patients and relatives attending a rehabilitation clinic, recruited through advertisements and leaflets. Each enrollee was provided with comprehensive materials, which included a cognitive impairment report and preventive methods against cognitive impairment based on the report, as a token of our appreciation for their time and commitment. Participants were included if they were aged older than 18 years, were fluent in Mandarin and signed informed consent. In order to compare the performance of the eQmci-CN and pQmci-CN, a diverse range of cognitive abilities were included. Exclusion criteria included: a history of dementia, stroke and/or nervous system diseases, active mental illness or delirium (including medication-related), the use of medications with known significant effects on cognition (e.g., antipsychotics, benzodiazepines, anticholinergics, analgesic opioids, chemotherapies, etc.), significant medical diseases interfering with cognition (e.g., OSAS, HIV, long COVID, etc.), history of alcohol or substance abuse, recent general anesthesia, significant visual or hearing impairment. Visual-spatial motor skills were assessed by asking participants to connect 4 points on a touchscreen by hand; those who failed to connect at least 2 out of 4 points were excluded. Hearing ability was screened by testing conversation bat a normal speaking volume, participants failed to do the question-answering conversation were excluded. Demographic information of each participant was collected including gender, age, education level, and occupation. All signed informed consent. This study received ethical approval from The Sixth Affiliated Hospital of Sun Yat-sen University (No. 2021ZSLYEC-297).

Measures

pQmci-CN description. The pQmci-CN was administered in its traditional pen and paper format, with an assessor recording the time and calculated the final score afterwards.

eQmci-CN description. The electronic Qmci-CN (eQmci-CN) was administered via iPad (iPad mini2, ME280CH/A) with the following subtest-specific adaptations: Clock Drawing Test (eCDT): Participants drew a clock showing ‘11:10’ using a stylus on the touchscreen. The system automatically quantified drawing latency (time from instruction to first stroke, reflecting executive function), digit position deviation (pixel-level accuracy of number placement), hand angle error (precision to 1-degree) timeout reminder < 60 s. ²⁸ Verbal Fluency: the electronic version of the Qmci uses the fruit category for verbal fluency assessment with real-time voice recording, while the paper version employs animals category. Logical Memory: Standardized audio playback of stories to eliminate examiner variability in delivery. All other subtests (Orientation, Registration, Delayed Recall) maintained the original Qmci-CN structure with touchscreen input replacing paper responses. The application would automatically skipped questions if no response was provided within the allotted time. The application automatically recorded the administration time. An assessor was available to address any participants questions. The interfaces are as showed in Figure 1A-C.

Figure 1.

The interfaces of introduction/basic information/question-answering in the eQmci-CN. (A) This is the introduction interface of the software for eQmci-CN. It lists various assessment items such as orientation, memory, clock - drawing test, delayed recall, fluency, and logical memory. There is also a warm reminder, stating that the assessment is expected to take about five minutes, and it is recommended to conduct the assessment in a quiet environment without distractions. For users who need to wear glasses, they are advised to prepare their glasses in advance. To ensure the accuracy of the assessment results, it is recommended to speak Mandarin and maintain a calm mood. There is a button labeled “Start Now” for users to begin the assessment. (B) This interface is used to collect the basic information of the participants. It requires users to fill in various details such as name, contact phone number, gender, height, weight, date of birth, educational attainment, occupation, whether there is a history of illness, whether wearing glasses, the main language used, and dominant hand. After filling in the information, users can click the “Confirm” button to proceed. (C) This is the question-answering interface. The instruction asks the user to listen carefully and respond after hearing the “ding” sound. There is a progress bar at the bottom, indicating the time remaining for the task, and a prompt at the bottom asking the user to speak their answer aloud.

Touchscreen experience index

The touchscreen experience data was gathered from each participant after completing the screening through self-reported five questions. These questions inquired about whether they used a touchscreen tablet and, if so, their average daily usage time; whether they used a touchscreen mobile phone and, if so, their average daily usage duration; as well as the degree of their familiarity with touchscreen devices. ²⁹ Regarding the scoring system, a binary scoring method was applied for the yes - no questions, where a “Yes” response was assigned 1 point and a “No” response was given 0 point. For the average daily usage, a five - level scoring scale was adopted: 1 point corresponded to usage of less than 1 h per day (“< 1h”), 2 points for 1–3 h per day (“1–3h”), 3 points for 3–5 h per day (“3–5h”), 4 points for 5–7 h per day (“5–7h”), and 5 points for usage exceeding 7 h per day (“> 7h”). The degree of familiarity was also scored on a five - point scale: 1 point for “Very unfamiliar”, 2 points for “Unfamiliar”, 3 points for “Uncertain”, 4 points for “Familiar”, and 5 points for “Very familiar”. Touchscreen experience index = Score of “Y/N for tablet use” + Score of “Average daily Tablet Usage” + Score of “Y/N for mobile phone use” + Score of “Average daily mobile phone Usage” + Score of “Degree of familiarity”.

Preference questionnaire

After completing the screening, they were asked to complete a validated 8-item questionnaire on their experiences of scoring the two versions of the Qmci-CN screen adapted from Wallace et al. ¹⁹ The questionnaire included items on comfort level, true performance representation, willingness to use an electronic version of a test in the future, negative experiences, acceptance and preference for paper versus electronic formats. Questions were scaled using a Likert scale from 1 to 7, where 1 = strongly disagree, 4 = neither agree nor disagree, and 7 = strongly agree.

Procedures

The tests were conducted at a clinic in the Sixth Affiliated Hospital of Sun Yat-sen University in Guangzhou, Guangdong, China or at participants’ homes. Whether in the hospital or at the participants’ homes, the assessment rooms were kept quiet, ensuring that the sound levels were consistently maintained below 50 dB.

Participants were randomly assigned to one of two groups: the pQmci-CN first group or eQmci-CN first group. A total of 4 administrators were involved in the study. These individuals consisted of experienced research assistants, and rehabilitation therapists. All administrators underwent rigorous training prior to the study. The training included detailed instruction on the proper administration of pQmci-CN and eQmci-CN, understanding the purpose and significance of each item, and how to handle potential participant questions or difficulties. They also participated in mock testing sessions to ensure proficiency and consistency in their approach. The two formats were administrated in a random order with 48-h washout by same administrators if possible. ³⁰ Once complete, participants were asked to complete the touchscreen experience index and questionnaire to evaluate their experiences with the two formats.

Statistical analyses

All statistical analyses were performed using SPSS version 25.0. The distribution of data was assessed using the Kolmogorov–Smirnov test, which indicated that most data were found to be normally distributed. Demographic data were compared using independent-samples t-tests. The Pearson correlation coefficient was used to examine the relationship between pQmci-CN and eQmci-CN scores for all participants. To compare the total and subscale scores between different groups based on administration order (pQmci-CN first versus eQmci-CN first) and administration modality, analyses of variance (ANOVA) was employed. Independent-samples t-tests were used to compare total and subscale scores between the pQmci-CN and eQmci-CN, while paired-samples t-tests were used to compare scores between modalities within and between first and second administrations within groups. Differences in administration times between pQmci-CN and eQmci-CN were analyzed using paired samples t-tests. Mann-Whitney U tests were used to assess the responses to the questionnaire. Demographic differences between preference groups (eQmci-CN versus pQmci-CN) were analyzed using independent t-tests. A level of statistical significance of 0.05 was used for all inferential analysis. Where appropriate, 95% confidence intervals (CI) were reported.

Results

Table 1 presents baseline characteristics of the total sample and univariate comparisons examining demographic and health-related variables for those who completed each of the two testing modalities first (i.e., those completing the pQmci-CN first and those completing the eQmci-CN first).

Table 1.

Demographic characteristics of all participants and according to order of administration of the two versions of the test.

	All participants [Mean (SD)] (N = 86)	pQmci-CN first admin [Mean (SD)] (N = 40)	eQmci-CN first admin [Mean (SD)] (N = 46)	p
pQmci-CN score	62.06 (11.69)	59.24 (12.86)	64.51 (10.08)	0.036†
pQmci-CN time (s)	287.41 (59.32)	294.15 (62.97)	281.54 (55.98)	0.329
eQmci-CN score	60.33 (12.52)	59.23 (14.19)	61.28 (10.94)	0.459
eQmci-CN time (s)	302.78 (51.54)	300.28 (52.6)	304.96 (51.09)	0.677
Gender [N (%)]				0.009*
Female	52 (60.5%)	30 (75%)	22 (47.8%)
Male	34 (39.5)	10 (25%)	24 (52.2%)
Age (y)	40.69 (15.67)	40.23 (15.35)	41.09 (16.11)	0.801
< 65	79 (91.9%)	37 (92.5%)	42 (91.3%)
≥ 65	7 (8.1%)	3 (7.5%)	4 (8.7%)
Occupation [N (%)]				0.792
Student	24 (27.9%)	12 (30%)	12 (26.1%)
Farmer	13 (15.1%)	6 (15%)	7 (15.2%)
Teacher	6 (7%)	2 (5%)	4 (8.7%)
Others	43 (50%)	20 (50%)	23 (50%)
Education (y)	13.5 (4.15)	13.28 (4.49)	13.70 (3.86)	0.642
< 6	10 (11.6%)	7 (17.5%)	3 (6.5%)
6–12	25 (29.1%)	8 (20.0%)	17 (37%)
> 12	51 (59.3%)	25 (62.5%)	26 (56.5%)
Smoker [N (%)]				0.114
Yes	9 (10.47%)	2 (5%)	7 (15.2%)
No	77 (89.53%)	38 (95.0%)	39 (84.8%)
Diabetes [N (%)]				0.160
Yes	2 (2.33%)	0 (0%)	2 (4.3%)
No	84 (97.67%)	40 (100%)	44 (95.7%)

CI: confidence interval; Qmci-CN: Quick mild cognitive impairment screen; pQmci-CN: paper version; eQmci-CN: electronic version. *p < 0.01; ^†p < 0.05.

A total of 95 participants were recruited for the study, all the participants pass the two entrance tests for visuomotor ability and hearing ability, 9 participants’ data were not included in the analyses because they did not complete both tests. There were no significant differences in demographic variables between those who completed the pQmci-CN first ( n= 40) and eQmci-CN first (n = 46) apart from gender; more females completed the pQmci-CN first (p = 0.009). The age of participants ranged from 18 to 82 years, with a mean education level of 13.5 years (range from 1 to 19y; standard deviation [SD] = 4.15). The total pQmci-CN scores ranged from 24 to 92/100 (M = 62.06; SD = 11.69), while eQmci-CN scores ranged from 27 to 88.5/100 (M = 60.33; SD = 12.52). Figure 2 presents the correlation between eQmci-CN and pQmci-CN scores for all participants. The Pearson correlation between participants’ eQmci-CN and pQmci-CN total scores was strong (r = 0.77) and statistically significant (p < 0.01).

Figure 2.

The correlation between individual participants’ paper Quick Mild Cognitive Impairment Screen (pQmci-CN) and electronic Quick Mild Cognitive Impairment Screen (eQmci-CN) scores.

Figure 3 shows the average score for both groups according to the version, highlighting that there was a moderate statistically significant correlation between the two groups according to their screening order (i.e., whether they scored the pen and paper version or electronic version first). This graph suggests that was a clear interaction between the modality (pQmci-CN or eQmci-CN) and the order in which each modality was presented, with the group scoring the eQmci-CN first showing a greater improvement in their scores when transitioning from scoring the eQmci-CN to the pQmci-CN than those completing the opposite screening order.

Figure 3.

Comparisons of Quick mild cognitive impairment (Qmci-CN) screen scores according to the order of modality first scored.

Table 2 shows the comparison between total and each subtest score according to whether the pQmci-CN or eQmci-CN were scored first. There were no statistically significant differences between groups except for registration (p = 0.023). For both groups, there were no statistically significant differences between the first and second administration total scores. For the pQmci-CN first group, the difference in scores was mainly attributed to the registration (p = 0.045), delayed recall (p = 0.003) and verbal fluency (p = 0.026). For the eQmci-CN first group, the difference in scores was attributable to differences in the registration (p = 0.001) and verbal fluency subtests (p = 0.007), which were statistically significant. The pQmci-CN used animal fluency, while the eQmci-CN used fruit fluency to avoid practice effects. Although animal fluency typically yields marginally higher scores (MD = 1.06 items in our data, p = 0.01), this difference is well within the expected range for semantically matched categories.

Table 2.

Comparison of scores for first test taken between two groups (n = 86) across total and subtest scores.

	pQmci-CN first [Mean (SD)] N = 40	eQmci-CN first [Mean (SD)] N = 46	95% CI of Difference of Means	p
Final score	59.24 (12.86)	61.28 (10.94)	−7.15,3.06	0.428
Orientation	9.73 (0.91)	9.91 (0.35)	−0.49, 0.12	0.223
Registration	4.45 (0.81)	3.96 (1.11)	0.07, 0.92	0.023†
Clock drawing	13.03 (10.99)	13.15 (3.10)	−3.75, 3.94	0.944
Delayed recall	15.90 (4.48)	16.39 (4.37)	−2.39, 1.41	0.609
Verbal fluency	8.11 (2.84)	7.20 (2.28)	−0.18, 2.02	0.101
Logical memory	11.03 (12.05)	10.67 (6.51)	−3.73, 4.43	0.865

CI: confidence interval; Qmci-CN: Quick mild cognitive impairment screen; pQmci-CN: paper Quick mild cognitive impairment. screen; eQmci-CN: electronic Quick mild cognitive impairment screen. ^†p < 0.05.

Test administration times, comparing the two modalities are shown in Table 3. The results show that the time to finish the pQmci-CN screen was significantly shorter than the eQmci-CN screening for the eQmci-CN first group (n = 46, p = 0.003) and all participants (n = 86, p = 0.005). The pQmci-CN was completed faster when administered after the eQmci-CN (MD = 12.61 s, p = 0.32, Cohen's d = 0.21), though this difference was trivial in magnitude. For the eQmci-CN itself, administration order had no discernible effect (p = 0.68, Cohen's d = -0.09).

Table 3.

Comparison of test administration times for the eQmci-CN and pQmci-CN.

	eQmci-CN [Mean (SD)]	pQmci-CN [Mean (SD)]	95% CI of Difference of Means	p
eQmci-CN first (N = 46)	304.96 (51.09)	281.54 (55.98)	−38.36, −8.47	0.003*
pQmci-CN first (N = 40)	300.28 (52.6)	294.15 (62.97)	−20.94, 8.69	0.408
All participants (N = 86)	302.78 (51.54)	287.41 (59.32)	−25.88，−4.87	0.005*

CI: confidence interval; Qmci-CN: Quick mild cognitive impairment screen; pQmci-CN: paper Quick mild cognitive impairment screen; eQmci-CN: electronic Quick mild cognitive impairment screen. *p < 0.01, ^†p < 0.05.

Touchscreen experience

According to the questionnaire, 90.7% of participants said they were “familiar” with (n= 47) or “very familiar” with (n = 31) the operation of the device before test administration (i.e., “I think my familiarity with the operation is?”). Only 4.7% of the participants (n = 4) rated this as “uncertain”. A “touchscreen experience index” was calculated using this information resulting in a total index score of 17/17. Overall performance on the eQmci-CN correlated moderately and significantly with the touchscreen experience index (r = 0.313, p < 0.0105), suggesting that more touchscreen experience (i.e., comfort with using a touch screen) was associated with higher eQmci-CN scores. There was no significant difference in touchscreen experience index scores according to whether participants scored the paper or electronic versions first (all items p > 0.05).

Preference questionnaire

After testing, participants completed a user satisfaction questionnaire. Participants’ preferences for the test modality (iPad versus pen and paper) were assessed on a 5-point Likert scale (1: strongly prefer paper, 5: strongly prefer iPad). Overall, 37.2% of participants slightly agree (n = 16), agreed (n = 5) or strongly agree d (n = 11) with the statement expressing their “Willingness to use electronic version of tests in the future”. In total, 23.3% (n = 20) of participants preferred the electronic version (scores 4–5), 38.4% (n = 33) had no preference (score 3), and 38.3% (n = 33) preferred the paper version (scores 1–2). Significant age differences were observed between groups (F = 2.95, p = 0.058), with pairwise comparisons showing clinically relevant trends ( Table 2 ). The electronic-preferring group was younger than the paper-preferring group (MD = 9.1 years, 95% CI [-1.23, 19.48], p = 0.095), though this did not reach conventional significance. No significant differences existed between other groups (all p > 0.1). Education levels did not differ across preference groups (p = 0.298). Mann-Whitney U tests indicated no significant differences between paper-first and electronic-first groups across all questionnaire items (all p > 0.05). Both groups showed comparable comfort with iPad assessments, acceptability of iPad quality, and test format preferences, with a tendency toward neutral or positive responses. Frustration rates were slightly higher for paper tests in the electronic-first group ( Table 4 ).

Table 4.

Mann-Whitney U tests of preferences questionnaire for two groups.

Questionnaire Item	Mann-Whitney U	Asymptotic p-value (2-tailed)
Comfort using iPad for assessment	870.0	0.654
Acceptability of iPad clarity/quality	974.5	0.626
iPad accurately reflects abilities	903.5	0.882
Need paper alternatives for some tasks	753.0	0.131
Willingness to use iPad in future	984.5	0.558
Frustration with paper-based tests	829.5	0.414
Frustration with iPad-based tests	844.0	0.497
Preferred test format	908.5	0.918

Scaled 1 through 7, 1 = strongly disagree, 7 = strongly agree, and 4 = neither agree nor disagree.

Discussion

This study compared the paper (pQmci-CN) and electronic (eQmci-CN) versions of the Qmci-CN in different age groups of cognitively normal subjects and their relatives to support the potential use of the eQmci-CN in clinical practice. The results showed that there was a strong and statistically significant correlation between eQmci-CN and pQmci-CN total scores (r = 0.77, p < 0.01), consistent with previous research examining other electronic version of cognitive tests such as the electronic version of the MoCA.^31,32 We also examined whether the order of screening would influence the performance of pQmci-CN and eQmci-CN and found that there were few differences in scores; only registration of the words (assessing working memory) was significantly higher for the pQmci-CN compared with the eQmci-CN (4.45/5 versus 3.96/5, respectively, p = 0.023). It might because the registration of the words is the second part that subjects need more time to adjust to the electronical test, while the first part of the test are random easy questions that do not need much effort on it. What's more, the absence of visual cues during auditory presentation (e.g., no lip-reading or textual support) may disproportionately impair older adults due to age-related declines in temporal auditory processing (reduced ability to resolve rapid speech) and cognitive load allocation (compensatory effort reduces working memory capacity). This aligns with age-related multisensory integration models, demonstrating that visual speech inputs robustly enhance auditory speech perception in older adults, especially under acoustically demanding conditions. ³³ There were significant differences in the administration times when the eQmci-CN was administered first, i.e., the eQmci-CN took significantly longer to score, albeit this difference was marginal. This indicates that the eQmci-CN might be influenced by the touchscreen experience when the eQmci-CN was administered first before the pen and paper version. This is supported by data showing that performance on the eQmci-CN correlated (r = 0.31, p < 0.01) with touchscreen experience index scores, i.e., more touchscreen experience was associated with higher eQmci-CN scores. However, this was only a marginal difference and would not be meaningful in clinical practice.

The analysis revealed important patterns in test performance related to administration sequence and digital literacy. “Touchscreen Experience Effects”: we observed a moderate but significant correlation between eQmci-CN performance and touchscreen experience (r = 0.313, p = 0.0105), confirming that digital familiarity influences electronic test scores. This supports the ‘touchscreen penalty’ hypothesis, where less experienced users may face additional cognitive load when navigating the digital interface. “Sequence-Dependent Effects”: the administration sequence produced markedly different outcomes: When eQmci-CN followed pQmci-CN, scores remained nearly identical (MD = 0.01 points), and when pQmci-CN followed eQmci-CN, we observed a clinically meaningful 3-point improvement. This pattern suggests two competing phenomena: “retest bonus”: ≈3-point learning effect from repeated exposure, and “touchscreen penalty”: ≈3-point interface adaptation cost.^20,34 In the pQmci-CN first group, these effects perfectly balanced (0.01 difference). In the eQmci-CN first group, the second test benefited from both the retest bonus and elimination of the touchscreen penalty, yielding the 3-point gain. These findings emphasize that while digital cognitive tests show good validity, clinicians should consider patients’ technological familiarity, maintain consistent test modalities for serial assessments, and allow for brief practice when transitioning to digital formats.

Although the 48-h washout period aligns with Qmci retest protocols, residual learning effects in cognitively normal individuals may require longer intervals (≥1 week) for optimal washout. ³⁰

The biggest difference in terms of subtest scores was seen for verbal fluency, which was found to have a statistically significant higher score for the pQmci-CN versus the eQmci-CN, irrespective of which was scored first. This may be explained by the different categories that were applied (i.e., the use of different versions of the test) such that verbal fluency for animals was used in the pQmci-CN and fruits were used in the eQmci-CN, in order to avoid learning effect. However, which semantic category is better to detect cognitive impairment is still under study.^35,36

That the test time of pQmci-CN was significantly shorter than that of the eQmci-CN (p < 0.05) has been found for other screening instruments including the eMoCA and its pen and paper version. ³¹ The reason why electronic versions may take slightly longer to administer may be due to the touch screen sensitivity or lack of user experience, especially among older people who are less technology savvy. ¹⁸ Practically, the administration of the electronic version comes with built-in strict subtest administration times, which cannot be overridden. Pen and paper versions despite timings, may be scored faster by trained raters as participants answer or indeed fail to answer sections. However, both versions demonstrated comparable administration times under five minutes—a practical advantage for busy clinical screening.^25,34 Notably, our non-randomized assignment of testing environments (clinic versus home) precludes analysis of potential setting effects on performance metrics. While this mirrors real-world clinical practice, future research should implement environmental randomization to better isolate contextual influences.

Demographic preferences revealed clinically meaningful patterns: electronic version preference was significantly associated with younger age (preferring group: 42.5 ± 12.1 years versus paper-preferring: 58.3 ± 15.7 years, p < 0.001), consistent with technology adoption trends. However, the substantial proportion (38.4%) expressing no preference suggests cross-generational adaptability of digital testing. Future implementations should include brief training for older adults to reduce digital literacy barriers.

The observed range of pQmci-CN/eQmci-CN scores (24–92 and 27–88.5) included a subset below clinical cut-offs, despite the sample's overall high education and youth. This variability likely reflects the intentional inclusion of individuals with subjective cognitive complaints, enhancing real-world applicability, and the tool's sensitivity to subtle differences, as evidenced by the absence of ceiling effects. This range supports the instrument's discriminative capacity, while significant correlations despite high baseline function underscore its robustness.

The electronic version's advantages include standardized timing and remote applicability, particularly valuable for rural populations with healthcare access disparities. During the recent COVID-19 pandemic, there has been increasing recognition of the need to develop more remote clinic activity. China is a big country with a very large population, in which the gap between urban and rural health resources is still widening such that people who live in relatively poor rural and remote areas still have limited access to advanced medical cares. ³⁷ Furthermore, given the ageing of the population, the burden of outpatient work is increasing. Thus, electronic cognitive screening could provide more clarity over those who should be referred for more detailed assessment in secondary and tertiary care. We suggest that based on its brevity and the existing validity of the pen and paper version, the eQmci-CN can meet the clinical needs and help identify those requiring specialized evaluation. ³⁸ Home assessment may also help overcome nihilistic views of cognitive decline and promote screening. According to the survey of World Alzheimer Report 2019, among approximately 70,000 respondents, 25% of them considered that there was nothing they can do about dementia. ³⁹ Further approximately 20% of those who are aware of their diagnosis of dementia decline to discuss it, because of fear over associated stigma. This has a significant and negative impact on these patients. ⁴⁰ Future research should compare the eQmci-CN with the pQmci-CN and other electronic versions of tests such as the eMoCA in formal diagnostic accuracy evaluation studies among patients referred for cognitive assessment. It would also be important to examine if such electronic versions administered in non-clinical settings could result in earlier detection of cognitive impairment and improve outcomes for patients.

Limitations

First, the study's sample size (N = 86), though adequate for preliminary validation, may limit the generalizability of findings across China's diverse demographic spectrum. While randomization was carefully implemented, the restricted sample, along with broad age ranges and high education levels, could amplify sampling errors and obscure subgroup differences. Future replication studies should employ stratified sampling to ensure proportional representation of regional, socioeconomic, and educational backgrounds. Second, as a pilot feasibility study, participant recruitment prioritized individuals with higher health technology engagement. Consequently, rural populations, who face the greatest healthcare access disparities, were underrepresented. This design limitation must be addressed in subsequent trials through targeted enrollment from medically underserved regions, where remote cognitive screening could yield maximal clinical impact. Third, although the eQmci-CN demonstrated acceptable performance in older subgroups, the limited sample of adults >65 years (particularly those with MCI/dementia) restricts the generalizability of clinical cutoffs. This aligns with phase 1 validation goals but necessitates confirmation in our ongoing elderly-focused study.

Conclusions

In summary, this is the first study to compare the electronic and paper versions of the Qmci screen, in this case the Chinese language version. The findings provide a foundation upon which to continue the development and validation of the eQmci as a cognitive screening tool. The development of online platforms for the eQmci-CN could further enhance its accessibility and usability. Although further study is required, it may become an effective tool for detecting prodromal cognitive impairment, supporting the potential future use of disease-modifying treatments to reduce the burden of dementia.