Introduction

The World Health Organization (WHO) estimating that depression impacted approximately 3.8% of the global population, equivalent to 280 million people of all age groups1. To implement effective intervention and prevention strategies, it is crucial to identify depression promptly and accurately. Utilizing measurement tools that provide a more comprehensive understanding of symptoms can facilitate the identification and lead to greater efficacy and accuracy in screening for depression2.

There are several commonly used assessment scales to measure depression, such as the Center for Epidemiological Studies–Depression scale (CES-D), the Hamilton Depression Rating Scale (HDRS) and the gold standard of self-rating scales Beck Depression Inventory (BDI-II)3,4. However, these scales tend to focus on physical and cognitive symptoms of depression and might not have enough items to evaluate the various dimensions of depression5. Interpersonal symptoms are a significant feature of depression and should be incorporated into assessment scales6. The interpersonal theory of depression, developed by Coyne7, proposes that depressed individuals contribute to a negative social environment through problematic relationships. Studies also suggested that different interpersonal stresses, social supports, and relational involvements, contribute to Asian adolescent depression8,9. Empirical studies reveal that depressed patients tend to have reduced social support and engage in more negative interactions with others10,11.

Cheung and Power12 reviewed the content of the 15 most widely used depression assessment scales indicating that interpersonal symptoms are often not given enough attention in these scales, with only a few of them containing up to three interpersonal items (Full scale displayed in supplementary material 1). Furthermore, many of these scales do not evaluate any interpersonal features of depression at all, with only 3% of interpersonal symptoms included in depression scales12. These statistics suggest that current depression assessment scales do not adequately consider the interpersonal aspect of depression and highlight the urgency of developing multidimensional measurement tools.

In light of the above, the 52-item Multidimensional Depression Assessment Scale (MDAS) was developed by Cheung and Power12, which consists of four sub-scales that cover almost the entire depressive symptomatology, including emotional (12 items), cognitive (16 items), interpersonal (12 items) and somatic (12 items). The interpersonal sub-scale emphasized features such as social withdrawal and problematic interpersonal relationships. Its multidimensional nature provides a holistic approach to measuring and evaluating depression severity. Numerous studies have demonstrated that the MDAS is a reliable and valid measure of depression, in both clinical populations and non-clinical ones12,13,14. Furthermore, it showed measurement invariance across cultures and genders, highlighting its flexibility for use in diverse populations13,15.

In the Asian population, depressive disorders are the leading contributor to the burden of mental disorders in Asia16. In the Asia region, adolescents make up almost 20% of the total population of the region17. Studies have shown that depression is the most common mental disorder among adolescents and its prevalence is increasing18. It is crucial for researchers to identify the factors and symptoms linked to depression in adolescents. Early detection and intervention can help prevent depression from developing into a formal diagnosis and persisting into adulthood, which can be a burden on society. Failing to address mental health issues in young people is also a serious public health concern and goes against Sustainable Development Goals 3.4 and 3.517.

Subsequently, the MDAS holds significant implications in assessing the interpersonal symptoms of Asian adolescents, research has shown that the MDAS is a reliable and effective screening tool for depression in high school-age adolescents, with measurement invariance across genders14. Asian cultures are heavily influenced by collectivism19. In these cultures, there is a strong emphasis on the collective “we” and fitting into society, leading to a focus on interdependence20. Therefore, assessing interpersonal symptoms of depression is significant for Asian populations12. In particular, Asian individuals are more likely to attribute the cause of their depression to interpersonal problems, and depressive symptoms aligning with collectivist interpersonal relationships and social functioning21.

However, the 52-item MDAS may face practical challenges due to its length, obstructing its use in large-scale health surveys for the preliminary diagnosis of depression. Additionally, it may be particularly difficult for individuals with depression who already have concentration difficulties22. Researchers are increasingly seeking valid and reliable measures that are concise despite tradeoff for poorer reliability and construct coverage23. Including too many questions in a survey can led to lower response rates24,25. Therefore, there is a need to develop a shorter version of the 52-item MDAS while maintaining its psychometric robustness.

Relentless scale shrinkage could compromise the psychometric properties of the scale26. Confirmatory Factor Analysis (CFA), a widely employed, classical test theory-based shortening strategy, emphasizes the selection of items through maximizing inter-item correlations27. Yet this approach may allow retaining of highly similar items which raises concerns regarding item redundancy.

To address the shortcomings of the popular methods, the current study utilized an alternative approach—Rasch analysis. The Item Response theory-based analysis examines the relationship between individuals and the items that measures a single construct or trait (e.g., depression). The probability of response across different items is modelled as a logistic function of the difference between the 2 parameter estimates: person location and item location. It assumes that the probability of given responses on items is governed by a person’s position on the underlying trait and the difficulty of the item as reflected in its location (the level of trait being measured). The likelihood of a higher score increases as individuals possess more of a certain trait and decreases as they exhibit less of the trait. This approach provides valuable insights into each item, facilitating the identification of redundant or irrelevant items28, thereby maintaining or even enhancing the accuracy of the shortened scale29,30. The specific objectives of this paper are twofold: (1) to shorten the original 52-item MDAS to a shorter version using Rasch analysis, and (2) to assess the psychometric properties of the shortened scale through Rasch analysis.

Methodology

Participants and procedures

A total of 1996 Asian participants aged 12–18 from secondary schools in the United Kingdom, Hong Kong, Taiwan, and the Netherlands were recruited using the Qualtrics online platform. Participants with limited literacy in written English or traditional Chinese were not included. As shown in Table 1, the final sample consists of 1996 Asian secondary students, with a mean age of 16.07 years and 58.7% of them were male. Hong Kong and Taiwan were the primary sources of Asian participants, accounting for 40.8% and 48.3%, respectively. The remaining 7.4% (147 participants) were from the UK, and 3.5% (70 participants) were from the Netherlands.

Table 1 Demographic variables.
Full size table

Statement of human and animal rights

The study obtained ethical approval from the ethical committee of the University of Hong Kong and was performed in accordance with the declaration of Helsinki.

Statement of informed consent and data confidentiality

Current study has ensured the informed consent from schools, parents, and students before collecting data. Participants and parents both provided their informed consent by checking a box on the online survey platform. Current study also ensured confidentiality and anonymity of participants’ responses throughout the research.

Measures

Depressive symptoms were assessed using the Patient Health Questionnaire (PHQ-9) and the original Multidimensional Depression Assessment Scale (MADS52-5pt).

Patient health questionnaire (PHQ-9)

The PHQ-9 is a nine-item self-report depression scale31. It corresponds to the DSM-IV criteria for depression, with responses scored on a Likert scale ranging from 0 (not at all) to 3 (nearly every day). The total scores obtained by adding the item scores indicate the severity of depressive symptoms, with higher scores indicating greater symptom severity. A cutoff score of ≥ 10 has demonstrated 88% sensitivity and specificity for major depression in the general medical population31. The PHQ-9 scale in the present study had a Cronbach’s alpha of 0.90.

The 52-item multidimensional depression assessment scale (MDAS52)

Developed by Cheung and Power12 the 52-item MDAS consists of a 5-point rating system. The psychometric properties of both the English and Chinese versions of MDAS52 are good. Both the English and Chinese version demonstrates high internal consistency (> 0.9) for the entire scale and sub-scale. The former shows significant and high convergent validity with the Beck Depression Inventory-II (BDI-II) in assessing depression symptoms and severity12. The Chinese version of MDAS52 correlates highly and significantly in terms of Spearman correlations between the MDAS, its sub-scales, the Edinburgh Postnatal Depression Scale (EPDS), and the Beck Depression Inventory (BDI), indicating good reliability and validity of the scale13,15. In the present study, the MDAS52 has a Cronbach’s alpha of 0.98.

Data analysis

Rasch analyses were conducted using the software – Winsteps 4.6.0 developed by Linacre and Wright32. The scale shortening was guided by testing of the psychometric properties of MDAS described in previous scale shortening using Rasch Analysis26,33. The parameters and their criteria that indicated the psychometric properties in the current study were summarized in the Table 2:

Table 2 Parameters and criteria in evaluating MDAS.
Full size table

The shortening procedures

Following the guideline, we started with examining the functioning of response categories through the monotonicity diagnostics45. After the appropriate number of response categories was confirmed, we launched the item removal by stepwise approach, with one item from each sub-dimension being removed in every subsequent round. The selection of items for removal was guided by their performance on the Rasch parameters. In each round, we first examined the parameters of items from all four sub-dimensional scales together in a general scale and removed those item that violates our established criteria.

If no issues were detected at the general scale level, we proceeded to examine each sub-dimensional scale individually to locate problematic items. If no anomalies were found, we selected items closest to being out-of-bound in the sub-scales as potential candidates for removal. We identified items with the highest and lowest fit MNSQ, the item with the greatest differential item functioning (DIF) contrast between genders, as well as item pairs exhibiting the largest residual correlations.

All potential candidates for removal were evaluated comparatively. The eventual readings on the 9 indices (item infit and outfit MNSQ, person infit and outfit MNSQ, item separation and reliability, person separation and reliability, and percentage of variance explained by the measure) from removing the proposed candidates were evaluated and ranked. We then discarded the item whose removal best preserved the 9 indices, as indicated by the calculated average rank, to maximize the number of removals that could be carried out. A candidate was rejected in comparison if its removal caused any of the 9 indices to fall outside the acceptable standard in Table 2. Furthermore, in cases where multiple items illustrated similar average ranks, priority was given to removing items that yielded better model fit.

After all dimensions had identified their candidates for removal in each round, we replicated the aforementioned steps to re-run the analysis for the general scale and then for the four sub-dimensional scales to detect deviated items in the next round.

The process terminated once the removal of an additional item caused any of the 9 indices in either the general scale or any of the sub-scales to fall outside the acceptable range. However, the item removal for the cognitive dimension was continued for four additional rounds to strive for a balance in the number of items in each dimension26.

Psychometric properties assessment

The psychometric properties of finalized MDAS scale were further assessed on statistical package SPSS (version 23.0) including the internal consistency reflected by the Cronbach’s alpha, concurrent validity assessed by Pearson correlation between the original scale and shortened scale, and the convergent validity by the Pearson correlation between the shortened scale and PHQ-9.

Results

Monotonicity diagnostics

According to Table 3, in the original 5-point MDAS design, each category had at least 4767 responses. Average ability measures and threshold values consistently increased across categories, with all outfit MNSQ values within the acceptable range. However, inadequate threshold distances were noted between categories 3 & 4 and 4 & 5, suggesting insufficient differentiation and prompting category collapse. After testing several collapsing patterns (3 & 4, 4 & 5, and 3, 4 & 5), the optimal configuration was achieved by collapsing categories 3 & 4, resulting in the best threshold distances and improved person separation index of 4.20.

Table 3 Monotonicity diagnostics.
Full size table

General scale

The preliminary analysis of the MDAS-52, as shown in Table 4, demonstrated satisfactory fit with infit and outfit MNSQ values of 1.00 logit for both items and persons. Most items fell within the acceptable MNSQ range of 0.7–1.3, except for item #28, which had values of 1.30 and 1.32. High reliability was observed for both person and item measures (≥ 0.95), and separation indices exceeded the criteria of 2.00 for persons and 3.00 for items. Minimal differential item functioning was noted between genders (DIF contrast < 0.43). In the test of unidimensionality, the PCA of Standardized Residuals revealed that the measures explained over 50% of the raw variance. However, items #7 and #8 were found to be locally dependent, with an excessive positive residual correlation (r = 0.21). Additionally, eigenvalues for the first (2.79), second (2.28), and third (2.02) contrasts exceeded 2.00 logits, indicating potential multidimensionality and item overlap.

Table 4 General Scale Readings.
Full size table

Eventually, the item removal terminated in the 6th round as any further item elimination would make the person separation of Somatic sub-scale falls below the threshold of 2.00, then the Cognitive sub-scale further went through 4 additional rounds of item removal to balance the item number in each sub-scale (totally 10 rounds of item removal). Afterward, the scale was reduced to 24 items (Detailed changes in General Scale Readings showed in supplementary materials 2). The final scale showed a slight decrease in person separation (4.20 to 3.46) and reliability (0.95 to 0.92), as well as item separation (6.28 to 5.54) and reliability (0.98 to 0.97). Despite these reductions, all item and person infit and outfit values remained at a perfect fit of 1.00 logit, and all items remained invariant across gender due to the elimination of misfitting items and the careful iterative removal of items to optimally preserve the scale functionality33.The identified issues from the preliminary analysis were resolved, with no items falling outside the acceptable MNSQ range of 0.7–1.3 or displaying excessive residual correlation. Additionally, unidimensionality improved, with the variance explained by the measures increasing from 52.1 to 54.1%, and the eigenvalues of contrasts falling below 2.00 logits.

Emotional sub-scale

Six rounds of item removal were conducted on the emotional sub-scale. Ultimately, items #3 (Low spirit), #5 (Sad mood), #6 (Guilt), #8 (Not cheerful), #10 (Dysphoric mood), and #11 (Shame) were discarded (Detailed item reduction process showed in supplementary materials 3). According to Table 5, in the final analysis of the emotional sub-scale, item and person separation dropped to 5.20 and 2.21, respectively, while item and person reliability decreased to 0.96 and 0.83. Item and person infit and outfit remained at 0.99, with a minor decrease of 0.01 in item infit. All items’ fit stayed within the 0.70–1.30 range and functioned consistently across genders as previous findings (largest DIF contrast was observed for item #12: 0.22). Positively, the percentage of variance explained by the measures increased to 57.4%, and the eigenvalues of contrasts remained below 2.00, ensuring the scale’s unidimensionality. However, 6 item pairs presented excessive negative residual correlations, with the highest of them as −  0.28 (#4 Gloominess & #9 Irritable mood, #4 Gloominess & #12 Anxiety).

Table 5 Emotional sub-scale readings.
Full size table

Cognitive sub-scale

In total, 10 rounds of removal were operated and items #14 (Loss of interest), #15 (No pleasure), #16 (The future feels bleak), #18 (Poor concentration), #20 (Life feels meaningless), #23 (Thoughts of suicide), #24 (Unable to make decision), #49 (Poor memory), #50 (Unable to plan things), and #51 (Feeling disorganized) were deleted (Detailed item reduction process showed in supplementary material 4). Table 6 below showed that the item and person outfit statistics remained at 0.99, with person infit slightly decreasing to 0.99 and item infit stable at 1.00 post-removal. Person separation and reliability declined to 2.08 and 0.81, while item separation and reliability dropped to 5.80 and 0.97. All cognitive items remained within the acceptable range of fit and functioned consistently across genders as previous findings (largest DIF contrast was observed for item #52: − 0.17). Although excessive residual correlation was observed among eight item pairs, with the highest at − 0.27 (#19 Self-blame & #52 Unable to care for oneself), PCAR results indicated good unidimensionality, with an increased percentage of variance explained by the measures and decreased eigenvalues of contrasts to below 2.00.

Table 6 Cognitive sub-scale readings.
Full size table

Somatic sub-scale

A total of 6 items, including #28 (Lower sex drive), #31 (Change in Weight), #32 (Crying), #34 (Slowed movement), #35 (More pain sensitivity), #36 (Intestinal problems) were discarded from 6 rounds of removal in somatic sub-scale (Detailed item reduction process showed in supplementary material 5). Finally, the item infit MNSQ of somatic sub-scale of MDAS-24 exhibited a drop from 1.00 to 0.99 while the else remained at 0.99, as shown in Table 7. Person separation and reliability were declined to 2.03 and 0.81, respectively, yet item separation grew slightly from 6.74 to 7.06 and item reliability kept the same. All items fit within the prescribed range, and DIF contrasts were below 0.43 (largest DIF contrast was observed for item #29: − 0.32), aligned with previous studies. Eight item pairs showed excessive negative residual correlations, with the highest at − 0.29 (#27 Change in appetite & #30 Fatigue). PCAR results indicated overall improvement, with the percentage of variance explained by the measures at 55.5% and the eigenvalue of the largest contrast was below 2.00 logits.

Table 7 Somatic sub-scale readings.
Full size table

Interpersonal sub-scale

Six rounds of item removal eliminated items #37 (Decrease in activities), #38 (Social withdrawal), #42 (Feeling undeserving of others care), #43 (Hypersensitive to criticism), #46 (Feeling let down by others), and #47 (Unable to love others) (detailed item reduction process shown in supplementary material 6). The final Rasch analysis (Table 8) demonstrated excellent fit, with item infit of 1.00, person infit of 0.99, and both person and item outfit MNSQ values of 0.99. Person separation and reliability decreased to 2.11 and 0.82, respectively, while item separation and reliability increased to 3.26 and 0.91. All items fit within the prescribed bounds (0.70–1.30) and functioned consistently across genders, with DIF contrasts below the ± 0.43 threshold for substantial bias41; notably, item #48 ("Aggression towards others") showed the highest DIF contrast of − 0.36, which, while below the critical threshold, warrants theoretical consideration. Most depression assessments (e.g. BDI) focus on internalizing symptoms without directly including aggression as a criterion. Yet a meta-analysis links irritability to depression (OR = 1.80), with 2%–5% of persistently irritable children developing later depressive issues46. Also in Asian adolescents, where collectivist cultures emphasize interpersonal harmony, assessing aggression can help capture culturally relevant relational disruptions47. Although six item pairs exhibited negative residual correlations approaching − 0.25 (highest at − 0.25 between #39 Feeling worse than others & #48 Aggression towards others), PCAR results supported robust unidimensionality despite keeping the above items which are conceptually important for depression assessment. The increased variance explained by the measures and all contrast eigenvalues below 2.00 indicated no significant alternative dimensions.

Table 8 Interpersonal sub-scale readings.
Full size table

Reliability test

A good internal consistency of the modified MDAS24-4pt was indicated by the high Cronbach’s alpha (alpha = 0.97), as well as every sub-scale: emotional (alpha = 0.89), cognitive (alpha = 0.89), somatic (alpha = 0.89), and interpersonal (alpha = 0.90).

Concurrent validity

The modified MDAS24-4pt preserved highly similar functionality as the original MDAS52-5pt. Significant high positive correlations were observed between the original MDAS52-5pt and the modified MDAS24-4pt (r = 0.98, p < 0.001), MDAS52-5pt-emotional sub-scale and MDAS24-4pt-emotional sub-scale (r = 0.96, p < 0.001), MDAS52-5pt-cognitive sub-scale and MDAS24-4pt-cognitive sub-scale (r = 0.95, p < 0.001), MDAS52-5pt-somatic sub-scale and MDAS24-4pt-somatic sub-scale (r = 0.94, p < 0.001), and MDAS52-5pt-interpersonal sub-scale and MDAS24-4pt-interpersonal sub-scale (r = 0.96, p < 0.001).

Convergent validity

Significant high positive correlations were observed between PHQ-9 and the modified MDAS24-4pt (r = 0.79, p < 0.001), emotional sub-scale (r = 0.70, p < 0.001), cognitive sub-scale (r = 0.74, p < 0.001), somatic sub-scale (r = 0.76, p < 0.001) and the interpersonal sub-scale (r = 0.75, p < 0.001).

Discussion

The major implication of the study was twofold. First, it is the first attempt to shorten 52-item 5 points MDAS (MDAS52-5pt) to a 4-point scale with 24 items (MDAS24-4pt) using Rasch analysis. Second, it evaluated the psychometric properties of the shortened MDAS24-4pt through Rasch analysis and additional Cronbach’s alpha and Pearson correlation tests. Nine parameters from RA, were examined in current research to facilitate the deduction of irrelevant and redundant items. Through ten rounds of item removal, the original MDAS52-5pt was condensed sharply to 24 items, with each sub-scale containing six items. And the MDAS24-4pt presents several improvements, aside from brevity, over the original scale in the following aspects:

Firstly, the monotonicity diagnostics revealed problems in the 5-point design of the MDAS52-5pt, where the threshold distances between the 3rd & 4th and 4th & 5th categories were not substantial enough to establish distinct categories. Consequently, categories 3 and 4 were collapsed to construct a 4-point design for the MDAS24, allowing the short scale to possess greater ability in distinguishing levels of severity of depression.

Secondly, the analysis detected a sex-related item #28 (Lower Sex Drive), showing underfitting, which entailed the item appearing less relatable with other items in measuring depression among adolescents, especially for Asians. The initial development of MDAS by Cheung and Power12 adopted a sample aged above 20 (Mean = 36.5, S.D. = 11.87). However, current study recruited 1996 Asian adolescents aged 12–18, with a mean age of 16.07. Also, a study revealed that Asians typically have their first sexual experiences later than other ethnicities, with a median age of 18.148. Therefore, the sexual dysfunction may not be a pertinent symptom for diagnosing depression in this demographic, and hence removed.

Next, an item pair exhibited excessive positive residual correlation, indicating redundancy, aligning with previous studies suggesting CFA’s tendency to retain highly similar items26. Through Rasch analysis, the overlapping items were identified and removed, resulting in no locally dependent items in the general scale after item reduction.

Further, the PCAR results of MDAS52-4pt indicated that three contrasts had eigenvalues exceeding 2.00 logits. However, after the item reduction, the eigenvalues of these contrasts fell below 2.00, indicating a reduction in alternative dimensions and the remaining items were more closely aligned in measuring a single underlying construct.

On the other hand, the drastic reduction in scale length inevitably leading to declines in certain aspects of functionality33. The 9 parameters in current study were therefore monitored throughout the process of item reduction to ensure the acceptable level of functionality was secured.

For example, we observed a decrease in person separation and reliability across all scales (general, emotional, cognitive, somatic, and interpersonal). These metrics indicate how well the scales distinguish between individuals with different ability levels. This decline reflects a common trade-off seen in other item reduction studies: as scales become shorter, their ability to discriminate between different ability levels often decreases26. As noted by Linacre41, fewer items inherently reduce the granularity of person ability estimates. Nonetheless, the final person separation (3.46) and reliability (0.92) exceed acceptable thresholds (≥ 2.00), ensuring the MDAS-24 remains effective for distinguishing levels of depression severity among Asian adolescents.

The reduction from MDAS-52 to MDAS-24 also resulted in several item pairs across sub-scales exhibiting negative residual correlations up to − 0.29, potentially suggesting suggesting a potential alternative dimension due to the altered scale structure, where endorsing one item reduces the likelihood of endorsing another49. However, these correlations remain below the threshold for significant local dependence adopted in many studies (within ± 0.3050or even ± 0.5051). The PCAR further supports the scale’s unidimensionality, with 54.1% of variance explained by the measures and all contrast eigenvalues below the threshold value of 2.0, indicating no significant secondary dimensions as according to the guideline by Linacre41. Additionally, the excellent overall model fit (infit/outfit MNSQ = 1.00) and strong individual item fit statistics indicate that the observed correlations reflect meaningful conceptual differences within the sub-scales, rather than problematic redundancy. Therefore, the minimal local dependence found in the MDAS-24 does not undermine its reliability or validity for assessing depression in Asian adolescents.

In general, despite the downsizing of item numbers and the associated fluctuation in scale functionalities, the eventual readings on Table 4, 5, 6, 7, 8 indicated that the finalized general scale and all sub-scales achieved the fairly acceptable functionality standards as displayed on Table 2. Additional tests also showed excellent psychometrics of the MDAS24-4pt, with high internal consistency from Cronbach’s alpha, high concurrent validity with the original scale, and strong convergent validity with the PHQ-9 from Pearson correlation tests. These findings support the MDAS24-4pt’s applicability for assessing depression among Asian adolescents.

The scale shortening also provided several clinical and research implications. First, it makes the MDAS a more cost and time-efficient, which may enhance response rates and reduce the burden on respondents. This brevity is particularly beneficial for organizations like educational units and primary health care conducting large-scale assessments on depressive levels among Asian adolescents26. In particular, the interpersonal symptoms of depression are associated with heightened vulnerability of interpersonal stressors during adolescence and a shorter depressive measure with interpersonal sub-scale would facilitate the measurement of a comprehensive profile picture in adolescents.

Next, by addressing the threshold issues and removing locally dependent and misfitting items, the MDAS24-4pt offers a more refined tool for diagnosing depression. The elimination of the sex-related misfitting item acknowledged the potential differences in depression symptoms across age groups and ethnicities, ensuring the scale’s relevance and sensitivity to its target demographic. The findings also highlight the necessity for future research to consider demographic-specific symptoms in the diagnosis of depression and other affective disorders.

However, the finalized scale presented several issues that necessitated further investigation. Firstly, although the scale was significantly reduced in length, it is imperative to examine empirical evidence regarding the exact degree of enhancement in the administration, such as a higher response rate, by the scale shortening.

Secondly, the current study detected item misfitting that could be attributed to the possible differences in the pathological symptoms across age groups or ethnicities. However, the current study recruited only Asian adolescents, making it unfeasible to perform validations across demographic groups.

Besides, previous studies have also suggested the possible variation in the depressive symptoms across different phases which may affect the scale validity and could be the direction for future study on MDAS24-4pt52.

Lastly, it is crucial to validate the shortened scale using a sufficient clinical sample to accurately establish the cutoff point for identifying individuals with depression34. The current study recruited Asian Adolescents merely from the dwelling population.

Conclusion

In conclusion, the study aimed to shorten the MDAS52-5pt scale to MDAS24-4pt using Rasch analysis and evaluated its psychometric properties. The downsized MDAS24-4pt showed declines in certain scale functionalities but still retained acceptable functionality standards. The MDAS24-4pt also presented improvements compared to the original scale, including reduced length, resolved monotonicity issues, identification and removal of misfitting and overlapping items, and minimized alternative dimensions, which enhanced the ease of administration and strengthened the measurement precision on Asian adolescents. However, further research is needed to investigate the impact of scale reduction, consider differences among demographic groups and different phases of depression, and validate the shortened scale using a clinical sample.