University Student Depression Inventory: Measurement model and psychometric properties

Abstract

University Student Depression Inventory (USDI) was developed to assess the symptoms of depression among the university students. Considering the debilitating nature of depression among university students globally, USDI was translated in Persian and validated using university students from Iran. A battery including the Persian version of USDI and scales measuring suicide, depression, and stress was administered to a normative sample of 359 undergraduate students, and an additional clinical sample of 150 students referred to the university's mental health centre. The results supported the factor structure and the psychometric properties of the translated version. Confirmatory factor analysis upheld the previously reported three‐factor first‐order and one‐factor second‐order structure. The internal consistency, test‐retest reliability, and concurrent and discriminant validity of the Persian version were supported. Cut‐off points using receiver operating characteristic curve analysis were established to identify students at risk. Gender differences on the symptoms of depression were evident only in the normative sample, where male participants, compared with female students, had higher mean scores in lethargy, cognitive/emotion, and academic motivation subscales. The translated scale can be used with Persian‐speaking students in Iran and the neighbouring countries as well as those settled in the West to identify symptoms of depression for further evaluation and management.

Key words:

• measurement model
• Persian version
• University Student Depression Inventory

University life is associated with many changes and challenges, which lead to the experiences of stress and depression (Bayati, Beigi, & Salehi, 2009; Verger, Guagliardo, Gilbert, Rouillon, & Kovess‐Masfety, 2010). Compared with the general population, previous studies have described particularly high levels of depression among university students (Adlaf, Gliksman, Demers, & Newton‐Taylor, 2001) because of new and unfamiliar circumstances, separation from family members, adjustment issues, lack of interest in their selected discipline, learning challenges, and academic stressors (Besharat, Rezazadeh, Firoozi, & Habibi, 2006). Depression is considered a serious problem among university students (Bayram & Bilgel, 2008), as depressed students report more academic difficulties than their peers (Vaez & Laflamme, 2008). Identification of these depressed students is essential for an effective management of their symptoms (Stallman, 2012).

Although there are a number of valid and effective scales to measure depression, most of the scales such as Beck Depression Inventory, Zung Depression scale, and Hamilton Depression Inventory measure clinical depression. It is argued that these previously existing scales measure dominant symptoms of depression while students' depression has been shown to be mostly a consequence of situational stressors (Romaniuk & Khawaja, 2013). The students' depression is marked by cognitive features such as low self‐esteem and perfectionism. They reflect emotional reactions and maladaptive behaviours such as apathy, lack of motivation to study, and disinterest in academic tasks (Das & Mishra, 2010). Most of the items in above mentioned established clinical measures severe depressive symptoms, such as a lack of appetite and sleep problems which may not be unusual to students given their lifestyle (Beck, Steer, & Brown, 1996).

University Student Depression Inventory (USDI) was developed using an Australian sample (for details, see Khawaja & Bryden, 2006) to measure the students' unique depressive symptomatology and to assist the mental health professionals working in university settings in accurate identification and management of vulnerable students. Item generation was based on the university students' and counsellors' perception and understanding of university student depression. Factor analysis resulted in a 30‐item scale that consisted of three subscales: lethargy (L), cognitive/emotional (CE), and academic motivation (AM). The L scale (nine items) focuses on fatigue and exhaustion, both mental (concentration difficulties) and physical (e.g., ‘I am more tired than I used to be’). The CE scale (14 items) targets the cognitive and emotional factors of depression, such as suicidal ideation, worthlessness, sadness, and emotional emptiness (e.g., ‘I feel worthless’). The AM scale (seven items) assesses motivation related to academic work (e.g., ‘I have no desire to attend lectures’). Overall, the subscales are aligned with the cognitive, emotional, behavioural, and motivational symptoms of depression (Das & Mishra, 2010). The USDI score ranges from 30 to 150, with higher scores indicating higher levels of depression. The reliability and validity of this instrument are well reported (Khawaja & Bryden, 2006; Romaniuk & Khawaja, 2013). Further, the stability of the factor structure is supported by a recent investigation on a large sample (Romaniuk & Khawaja, 2013). Recent cross‐cultural investigations have indicated that USDI can be used in non‐Western cultures and has the potential of being used in universities outside Australia (Khawaja, Santos, Habibi, & Smith, 2013). One of the countries of interest is Iran because of its emphasis on education and training and a substantially large population of university students.

In Iran, a vast majority of the population is literate and speaks the Persian language. Persian is also spoken in Iran's neighbouring countries, such as Tajikistan and Afghanistan (Joharifard, 2010). Since the 1979 revolution, the country's emphasis on education has increased, and there are 93 universities in the country and, currently, 5.4 million university students in Iran (Iranian National Organization of Educational Testing, personal communication, 2013). In recent decades, Iran has experienced dramatic social change, including significant emphasis on education and social activity among youth, especially females (Abbasi, Mehryar, Jones, & McDonald, 2002). Women make up 65% of the university students in Iran. In general, Iranian youth are under a strong pressure from parents, educational institutions, and the society at large to achieve academically, to develop future competencies, and to compete internationally. These demands have led to an increase in academic stress and depression (Shokri et al., 2008).

An emphasis on education and well‐being of the students has initiated a keen interest in the research on university students in Iran. Even though the research is sparse, it suggests that a considerable percentage of Iranian students exhibit varying levels of depression or are at risk of becoming depressed, with an overall depression rate of 8.5–44% (Amini & Farhadi, 1999; Ebrahimi & Keyghobadi, 2004; Rahimi & Kamran‐Pour, 2006). This broad range of prevalence of depression is mainly due to different types of scales and methodologies used by the researchers. Further, contrary to the data emerging from the West where depression is more prevalent among female than male students (Khawaja & Duncanson, 2008), the results are mixed in Iran. Although depression is reported by female students in Iran, the rate is not as high as in males, who have emerged to be more vulnerable than females because of the social and familial pressure to earn money and to support the family financially (Bayani, Ghodarzi, Bayani, & Kochaki, 2008). Prior studies in Iran have shown that university education is a more stressful event for males than for females (Dehshiri, Borjali, Sheikhi, & Habibi, 2008; Khodayari‐Fard, Shokohi‐Yekta, & Ghobari, 2004), and subsequently, these stressors are associated with depression (Khodayari‐Fard et al., 2004; Rezai‐Adriany, Azadi, Ahmadi, & Azimi, 2007). However, some researchers have found only non‐significant differences based on gender (Jahani, Norozi, Hasan‐Pour, Shamlo, & Sarichlo, 2008). The unclear gender‐based prevalence is due to the different assessment tools used by the researchers (Ghasemzadeh, Mojtabai, Karamghadiri, & Ebrahimkhani, 2005). The depression rate and gender‐based prevalence are still unclear in Iran and warrant further investigation (Mellsop & Smith, 2007).

Keeping in view that depression appears to be a major health problem in Iranian university education, it is necessary to have a valid and reliable measure tailored for this population. Currently, some of the other significant measures such as Beck Depression Inventory, Zung Depression Scale, Hamilton Depression Inventory, Symptom Checklist‐90, and General Health Questionnaire are available in Persian language. However, as stated earlier, they are more appropriate for clinical settings (Martin, Swartz‐Kulstad, & Madson, 1999) and may not be appropriate for university students (Khawaja & Bryden, 2006). Therefore, development of a specific scale for Iranian student population seems necessary. Further, there is a need to adapt and translate the scale in a language appropriate for this cultural group. Moreover, it is important to investigate the depression in Iranian students in order to understand how depression is manifested in this population compared with the youth living in other nations and cultures. Thus, such investigations would also help with future predictions and management of depression among Iranian students.

The literature review indicates that in spite of the global nature of depression, the bulk of the previous research has examined depression among Western student populations. While the original USDI has been developed in Australia (Khawaja & Bryden, 2006), no published psychometric studies of the USDI in non‐Western cultures are known to the authors. The Persian edition is expected to find widespread use in various settings in Iran, other Persian‐speaking countries, and among practitioners working with Persian‐speaking clients in western countries. Although the English version of the USDI has satisfactory psychometric properties, to date, no Persian‐language translation has been conducted. Further, no study has tested the construct validity of the USDI in a sample drawn from Iran. The present study aimed to translate the USDI into the Persian language and validate it for Iranian students. It was expected that the instrument's factor structure and psychometric properties would be upheld with the Iranian student population. Further, it was expected that the Persian version would reveal a possible gender difference. The study also aimed to develop cut‐off points for the USDI so that it can be used as a screening tool to identify Iranian students who may require a full clinical examination.

Method

Participants

The study participants were grouped into two samples: general students (normative sample) and students who had been diagnosed with depression (clinical sample). Although 400 undergraduate students volunteered to participate in the normative sample, 41 (10%) students were excluded because they did not meet the inclusion criteria. Participants were considered eligible if they were within the age of 17–40-years, had at least completed one semester of university education, and provided written informed consent for participation in the study. Those who met the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition, Text Revision (DSM‐IV‐TR) criteria of psychiatric disorders such as psychoses, organic brain disorder, and/or drug dependence were excluded from the study. Thus, the participants were comprised of 359 undergraduate students (206 male and 151 females; data missing for two participants) from Tehran University. Their average age was 21.26-years (standard deviation (SD) = 2.60; range 17–33). Majority of the students (n = 257) were single, while some (n = 32) were married, one was divorced, and 69 participants did not specify their marital status. Most of them (n = 187) lived in dormitories, while other (n = 72) lived with their families, or in rental accommodations (n = 100).

The clinical sample consisted of 150 students (70 males and 80 females) who had sought help, as outpatients, at the university health clinic. Psychiatrists, using DSM‐IV‐TR‐based interviews, had diagnosed 79 with major depression, and the others were regarded as non‐depressed. The average age of students was 21.46-years (SD = 2.40; range 17–29). Majority of them (n = 125) were single, 22 were married, and 3 did not specify their marital status. Nearly half (n = 76) lived in dormitories, some of them lived with their families (n = 60), or in rented accommodations (n = 14).

Measures

USDI

The USDI is a 30‐item self‐report rated on a 5‐point scale (1: Not at All to 5: All the Time) (Khawaja & Bryden, 2006). The three subscales are lethargy (L), cognitive/emotional (CM), and academic motivation (AM). Its Cronbach's alpha was α = 0.95, and the internal consistencies of the subscales of L, CM, and AM were 0.89, 0.92, and 0.84, respectively. The correlation coefficient for its test‐retest reliability with a 1‐week interval was 0.86. The scale had good concurrent validity with a strong positive relationship with the Depression, Anxiety, and Stress Scale (Khawaja & Bryden, 2006). Romaniuk and Khawaja (2013) confirmed the original three‐factor model with an additional second order total score factor as a good fit to the data.

Suicide Ideation Scale (SIS)

The SIS consists of 38 items, rated on a 3‐point scale (1: Not at All to 3: Very Much). It was developed in the Persian language for the student population (Mohammadifar, Habibi, & Besharat, 2006). An exploratory factor analysis (EFA) revealed five subscales, labelled as feelings of guilt, hopelessness, withdrawal, inertia, and depression. The internal consistency for the total score and subscales ranged from 0.72 to 0.93; the 2‐week test‐retest reliability for the total and subscales ranged from 0.81 to 0.89. The Pearson correlation of the total score of SIS with the Beck Hopelessness Scale was −0.31 (p < .001).

Beck Depression Inventory Second Edition‐Persian version (BDI‐II)

The BDI‐II consists of 21 self‐report items—each item has four statements. It is used to measure depression and is comprised of general depression and hopelessness, emotional distress, negative attitude, and psychosomatic disorder factors. The BDI‐II's internal consistency and 1‐week test‐retest reliability among younger adult outpatients were 0.92 and 0.93, respectively (Beck et al., 1996). For Iranian students, the Cronbach's alpha, test‐retest reliability, and its correlation with Minnesota Multiphasic Personality Inventory‐2 (MMPI‐2)'s depression scale was 0.87, 0.49, and 0.60, respectively (Rajabi, Attari, & Haghighi, 2001).

Student‐life Stress Inventory (SSI)

The SSI (Gadzella, 1994) assessed university students' life stressors and reactions to stressors. The scale is comprised of 51 items measured on a 5‐point scale (1: No, Never to 5: Most of the Time). Internal consistency, as indicated by coefficient alpha, ranged from 0.63 to 0.92 for subscales (Gadzella & Baloglu, 2001). A Persian version of SSI (Shokri et al., 2008) was used in the current study.

Procedure

USDI was translated into Persian and back‐translated by a research team comprised of linguistics and psychologists fluent in both languages. Every effort was made to ensure that the translated version conveyed proposed meanings in Persian. Translation was completed using the guidelines for the cross‐cultural adaptation of instruments (Guillemin, Bombardier, & Beaton, 1993). Differences were judged and resolved based on the consensus reached by the mental health professionals in the research team. Two other mental health professionals completed back translation. Again, differences were resolved by agreement, which led to the final version. The Persian version was trialled on 30 students, who confirmed a comprehensive understanding of the scale. Apart from a few minor adjustments to the wording and layout, the Persian version was similar to the original USDI.

The ethical approval was obtained from the university's research committee. The study was advertised on the university campus, and students were invited to participate. They were informed that their participation was voluntary and that they could discontinue at any time. Participants were also informed about confidentiality. Out of the total normative sample, 30 males and 30 females completed the USDI twice with a 4‐week interval for test‐retest reliability purposes.

Results

Analysis strategy

The data were cleaned and screened. The assumptions of normality were checked, and a slight skew was evident in the subscales but not in the total USDI score in normative group. A decision was made not to transform the data because the dataset were large and the transformation did not improve the results. The decision to whether remove or retain the outliers was made by comparing the original mean with the 5% trimmed mean (Tabachnick & Fidell, 2013). The Cronbach's alphas for SIS, BDI, and SSI were 0.88, 0.85, and 93, respectively, indicating satisfactory internal consistency of these measures.

Confirmatory factor analysis was selected to examine the USDI's stability. This method offers a variety of statistical tests and indices designed to assess the ‘goodness‐of‐fit’ of the identified models (Maccallum, Browne, & Sugawara, 1996). In the present study, the goodness of fit was evaluated using the following statistics: the non‐normal fit index (NNFI > 0.90), the comparative fit index (CFI > 0.90), normal chi‐square (3 > χ²/df < 2) and the root mean square error of approximation (RMSEA) and its 90% confidence interval (CI) < 0.05 (Miles & Shevlin, 2007). Multiple indices were used because they provide different information about the model's fit (i.e., absolute fit, fit adjusting for model parsimony, fit relative to a null model). Together, these indices provide a more conservative and reliable evaluation of the solution (Maruyama, 1997). Because of multivariate skewness in the data, the fit indices of all the models were corrected with the Satorra‐Bentler scaled difference chi‐square test statistic (Bentler, 1995; Hu, Bentler, & Kano, 1992).

The fitted models were nested; in these instances, the comparative fit was evaluated by chi‐square difference tests (Δχ²) and the interpretability of the solutions. The concurrent validity was investigated by examining the correlations between the USDI scores and SIS, BDI‐II, and SSI. To evaluate the test‐retest reliability of the USDI, Spearman correlation coefficients were calculated at two points of time over 4-weeks for the total scale and three subscales. Cronbach's alpha and mean inter‐item correlation coefficients were calculated for the total USDI and its subscales. To explore the relationship between the USDI and the remaining measures, the Spearman r correlation was used to address the skewness of the scores. Given the number of correlations, the p‐values were set at 0.012 to control for the experiment‐wise error. The Bonferroni adjustment was used: An initial Cronbach's alpha of 0.05 was divided by the number of measures or 0.05/4 (Tabachnick & Fidell, 2013). Finally, to develop cut‐off points, the receiver operating characteristic (ROC) curve analysis method was used (Beck & Shultz, 1986).

LISREL version 8.72 (Jöreskog & Sörbom, 2005) was applied to the current data to examine the fitness of four models. Model 1 describes a one‐factor model in which all 30 items were made to load on a single factor of general depression symptoms. Model 2 presents a three‐factor orthogonal model, Model 3 examines a three‐factor oblique model, as reported by Khawaja and Bryden (2006) for the EFA procedure, and Model 4 examines a three‐factor oblique first‐order and single‐factor second‐order model, as suggested by Romaniuk and Khawaja (2013). The oblique model was used because we expected the factors to be theoretically correlated. For all the models, the variance of each factor was set to a 1.0-Z score for the univariate skewness values ranging from −0.37 (Item 21, ‘My mood affects my ability to carry out assigned tasks’) to 6.58 (Item 7, ‘I have thought about killing myself’; Table 1). Thus, we used the weighted list square because of its lower sensitivity to normality (Bentler & Bonett, 1980). We used PRELIS (Jöreskog & Sörbom, 2005) to estimate the polychoric correlations and their asymptotic covariance matrix of the sample variance and covariance.

Table 1. Parameter estimates and goodness‐of‐fit indexes for CFA of the USDI

Items	P.E1	Z	P.E2
Lethargy
1. I am more tired than I used to be	0.81	0.96	0.68
4. I do not have the energy to study …	0.77	0.50	0.67
9. My energy is low	0.91	1.66	0.77
13. I find it hard to concentrate	0.89	0.98	0.77
16. I don't feel rested even after sleeping	0.85	2.21*	0.73
18. Challenges I encounter in my studies …	0.76	1.76	0.68
21. My mood affects my ability to carry …	0.49	−0.37	0.42
24. Daily tasks take me longer than they …	0.73	1.47	0.67
28. My study is disrupted by distracting …	0.87	0.36	0.69
Cognitive/Emotion
2. I wonder whether life is worth living	0.77	3.41**	0.64
5. I feel worthless	0.91	3.80**	0.73
7. I have thought about killing myself	0.66	6.58**	0.59
10. No one cares about me	0.79	2.91**	0.67
11. I feel emotionally empty	0.86	1.82	0.73
14. I feel sad	0.87	0.67	0.79
15. I worry I will not amount to anything	0.72	1.68	0.62
17. The activities I used to enjoy no …	0.72	1.08	0.66
19. I feel like I cannot control my …	0.62	2.44	0.59
20. I spend more time alone than I used to	0.87	0.84	0.69
22. I feel disappointed in myself	0.95	1.51	0.81
25. I feel withdrawn when I'm around …	0.78	2.97**	0.69
26. I do not cope well	0.71	1.68	0.71
29. I think most people are better than me	0.57	1.56	0.50
Academic Motivation
3. I do not have any desire to go to lectures	0.73	0.85	0.51
6. I don't attend lectures as much as I …	0.89	1.32	0.64
8. I don't feel motivated to study	0.91	1.93	0.73
12. Going to university is pointless	0.92	4.06**	0.66
23. I have trouble starting assignments	0.93	0.69	0.76
27. I do not find study as interesting as …	0.97	2.98*	0.74
30. I have trouble completing study tasks	0.87	1.24	0.77
Model	NNFI	RMSEA	CFI	S‐B χ^2	Δ S‐B χ^2
M1 = one‐factor model	0.90	0.12 (0.115–0.124)	0.90	2486.90	1221.65^1***
M2 = three‐factor orthogonal model	0.96	0.077 (0.072–0.082)	0.96	1265.25	229.35^1***
M3a = three‐factor oblique model	0.97	0.066 (0.061–0.071)	0.97	1035.90	282.32***
M3b = three‐factor oblique model and correlated errors	0.97	0.051 (0.046–0.057)	0.97	753.58	29.07***
M4 = three‐factor oblique model, correlated errors and second order total score	0.98	0.049 (0.044–0.055)	0.98	724.51

Notes. CFA = confirmatory factor analysis; P.E₁ = parameter estimation for three‐factor oblique model; P.E₂ = parameter estimation for three‐factor oblique model, correlated errors, and second order total score; Z = Z score for tests of univariate normality; *p < .05, **p < .01, ***p < .001.

Confirmatory factor analyses

Table 1 represents the fit estimates for all models. The one‐factor model and the three‐factor orthogonal model did not meet the previously specified fit criteria, and the three‐factor oblique model showed inadequate fit to the data (M₁–M₃). Although modification by freed errors in the three‐factor correlated model revealed some improvement, it still did not meet all criteria (M_3b; RMSEA > 0.05). The chi‐squared test was significant for all the models, but that is to be expected for models with large degrees of freedom (df) and relatively large sample sizes (Bentler, 1995). An examination of the remaining fit indices suggested that the Satorra‐Bentler scaled chi‐square statistic (Satorra & Bentler, 2001) to find the best fitting model showed that M₄ (Table 1) was significantly better than the three‐factor oblique and correlated errors model (Δχ² = 29.07; p < .001). In a comparison of the nested models, the Δχ² showed that the three‐factor oblique, correlated errors, and single‐factor second‐order model provided a better fit (S‐B χ²/df = 1.87; CFI = 0.98; NNFI = 0.98; and RMSEA = 0.049 ((CI) 90% = 0.044, 0.055)). The correlation between the L and CE latent variables was 0.74 (p < .001); between L and AM, it was 0.46 (p < .001); and between CE and AM, it was 0.70 (p < .001). Because of the high correlations between the three latent factors, the possibility of a second‐order depression factor was tested. The correlation between the L, CE, and AM first‐order latent variables with depression factors was 0.56, 0.80, and 0.44 (p < .001).

Concurrent validity

Table 2 shows the relationship between USDI, BDI‐II, SIS, and SSI. Concurrent validity was indicated by USDI's positive relationship with BDI, SSI, and SIS. The strength of relationships for all coefficients was moderate or high.

Table 2. Spearman correlations of USDI and subscale scores with BDI, SIS, and SSI based on normative sample

Measures	SIS	BDI	SSI
Lethargy	0.66**	0.72**	0.63**
Cognitive/Emotion	0.66**	0.79**	0.68**
Academic Motivation	0.47**	0.62**	0.51**
USDI	0.65**	0.75**	0.69**

Notes. **P < .01 (two‐tailed), SIS = suicide ideation scale, BDI = Beck's Depression Inventory, SSI = Student‐life Stress Inventory, and USDI = the University Student Depression Inventory.

Consistency and test‐retest reliability

Table 3 presents the mean, standard deviation, internal consistency coefficients, and mean inter‐item correlation of the USDI based on sample type (normative/clinical). The test‐retest reliability of the USDI according to the Spearman correlation coefficients for L, CE, AM subscales, and total score was 0.86, 0.72, 0.69, and 0.80, respectively. This was moderately high. The means of inter‐item correlation were 0.44, 0.45, 0.49, and 0.36 for L, CE, AM, and USDI, respectively. In the inter‐item correlation range between 0.2 and 0.4 and the corrected item‐total correlation, all the items performed adequately (range of 0.37–0.75). In this case, the Cronbach's alpha coefficients for the USDI total scores were 0.94 and 0.83 in the normative and clinical groups, respectively.

Table 3. Means, standard deviations, internal consistency coefficients, and mean inter‐item correlations of USDI

	Lethargy		Cognitive/Emotion		Academic motivation		USDI
Normative samples (n = 359)
Sample	Male	Female	Male	Female	Male	Female	Male	Female
M	24.39	21.84	33.13	29.80	17.82	15.72	75.34	67.37
SD	7.63	6.85	10.63	11.74	6.80	6.25	21.26	20.94
α	0.88	0.87	0.90	0.94	0.85	0.89	0.94	0.95
Mr.	0.44	0.42	0.38	0.53	0.46	0.53	0.33	0.37
Clinical samples (n = 150)
M	26.73	27.08	40.43	38.90	20.69	19.56	88.49	86.35
SD	7.83	8.16	12.01	15.16	8.91	8.58	20.94	26.10
α	0.92	0.91	0.94	0.96	0.94	0.95	0.93	0.96
Mr.	0.55	0.53	0.53	0.67	0.71	0.73	0.31	0.44

Notes. α = Cronbach's alpha; M = mean; Mr. = mean inter‐item correlation.

Gender differences

Table 3 shows the means and standard deviations of USDI and its subscales for males and females, separately, in both clinical and normative groups. In the normative group, the male students scored significantly higher than the females on their total USDI scores (t(355) = 3.52, p < .001). The male students in the clinical group also scored slightly higher than the females on their total USDI scores; however, the difference did not reach a significant level (t(148) = 0.21, p = .83, ns). In addition, a multivariate analysis of variance (MANOVA) was conducted to investigate the gender‐based difference between males and females in the three USDI subscales (as dependent variables) with gender used as an independent variable in the analysis. The Box's M assumption of the homogeneity of variance–covariance matrices was violated in the normative group, (F_normative (6, 720162.8) = 2.44, p < .05) and (F_clinical (6, 48408.35) = 1.00, p = .42). However, Box's M is considered a notoriously sensitive test, while MANOVA is robust to violations of its homogeneity of variance when the sample sizes are large (Tabachnick & Fidell, 2013). Gender also had a significant effect on the USDI subscales in the normative sample: Hotelling's Trace F_normative (3, 353) = 4.82, p < .01, partial Eta squared = 0.04; and Hotelling's Trace F_clinical (3, 146) = 0.28, p = .84, ns, partial Eta squared = 0.01. This effect was observed univariately on the USDI subscales. In the normative sample, the males scored significantly higher than the females on the L, CE, and AM subscales: (F(1, 355) = 10.54, p < .01), (F(1, 355) = 7.82, p < .01), and (F(1, 355) = 8.85, p < .01), respectively. In the clinical sample, there was no significant difference between males and females on the L, CE, and AM subscales: (F(1, 148) = 0.12, p = .73, ns), (F(1, 148) = 0.08, p = .78, ns), and (F(1, 148) = 0.20, p = .66, ns), respectively.

Cut‐off scores

A ROC curve that plots sensitivity versus specificity for every possible cut‐off point was obtained. Youden's index was used to evaluate the optimal cut‐off point (sensitivity + specificity—1.00) (Viinamäki et al., 2003). Sensitivity and specificity indices were calculated for all the possible USDI cut‐off points. The ROC curve was calculated to estimate the instrument's discriminant capability. USDI raw scores were analysed to classify both at‐risk and non‐risk groups. The best USDI cut‐off point for females is 91 with a sensitivity of 85.37% and a specificity of 84.62%, indicating that 15.38% of the non‐risk group and 85.37% of the at‐risk group exceeded the cut‐off of 91. The area under the curve was 0.89 ((95% CI) = 0.79–0.95, p < .001). The best USDI cut‐off point for males is 85 with a sensitivity of 92.11% and a specificity of 78.12%, indicating that 21.88% of the non‐risk group and 92.11% of the at‐risk group scored beyond the cut‐off of 85. In this case, the area under the curve was 0.91 ((95% CI) = 0.82–0.97, p < .001). These cut‐offs were applied because they were revealed to be the most optimal combination of the points in the sensitivity and specificity indices (Perkins & Schisterman, 2006). Comparisons of the area under curve (AUC) in both sexes showed that no significant difference in the AUC is associated with gender (d = 0.03, standard error = 0.05, Z = 0.54, p = .59, ns).

Discussion

The present study examined the factor structure, psychometric properties, and clinical utility of the Persian version of USDI. The results were consistent with past investigations on USDI. The Iranian university students endorsed the physical, cognitive, and emotional manifestations of depression observed previously in the university students residing in the West (Khawaja & Bryden, 2006; Romaniuk & Khawaja, 2013). Therefore, results support the universal nature of depression. However, gender differences were found, according to which the male students in Iran tend to be more depressed than the female students. In general, the translated scale emerged as a robust measure that can be used with Persian‐speaking university students to identify those at risk and warranting further evaluation and intervention.

The USDI's original three factors were upheld in the Persian version. Consistent with the earlier investigation (Khawaja & Bryden, 2006), the three factors of the Persian scales were correlated. Further, the findings of this study corroborate with the recent analyses (Romaniuk & Khawaja, 2013) as the three dimensions, merged to form a second‐order factor reflecting the overall depression of the university students. It is interesting to note that the Iranian university students experienced lethargy and physical exhaustion (L), a range of cognitive and emotional difficulties (CE), and a low motivation to attend lectures and other academic tasks such as assignments and readings (AM) (Bayati et al., 2009; Shokri et al., 2008). The findings indicate that the stressors and reactions of university students tend to be similar across the globe.

The translated USDI is identified as an internally consistent scale. The Cronbach's alpha coefficients in the normative and clinical groups were above 0.7, and the item‐total correlations exceeded the minimum acceptable value of 0.30 (Nunnally & Bernstein, 1994). In the present study, all the items performed adequately (within a range of 0.37–0.75). Therefore, the results supported the internal consistency of the overall scale and its subscales, and the items within each subscale were thematically linked.

In general, the validity of the full scale and the subscales of the Persian USDI were supported. Consistent with previous research, it also appeared that experience of depression impacted negatively the academic achievement of the Iranian students (Bagheri‐Yazdi, Bolhari, & Peiravi, 1995). Moreover, the full scale and its subscales positively correlated with other scales measuring depression, stress, and suicidal ideation (Khawaja & Bryden, 2006; Romaniuk & Khawaja, 2013). The discriminant ability of the scale was further evident by cut‐off scores that could be used to identify depressed students at risk. Future mental health professionals who offer services in the university clinics would benefit from these cut‐off scores.

The findings showed the male students to be more depressed than the female students in the normative groups. Similar trends were revealed in the clinical group. It is possible that these differences may be related to cultural issues (Goldberg, Oldehinkel, & Ormel, 1998; Sen & Mari, 1986). Consistent with previous research, cultural expectations regarding gender roles in the society have profound effects on mental health in general and depression in particular (Kokanovic, Dowrick, Butler, Herrman, & Gunn, 2008). It is important to note that limited job opportunities after graduation would have more severe effects on male student than female students, who are expected to bear the financial burden of the family (Bayati et al., 2009). Therefore, these stressors and social pressures make the university education more stressful event for males than for females (Dehshiri et al., 2008; Khodayari‐Fard et al., 2004; Rezai‐Adriany et al., 2007). It is possible that these gender differences may be due to different ways of coping with stressors (Pillay & Ngcobo, 2010) or due to different help‐seeking behaviours (Mellsop & Smith, 2007; Tapsell & Mellsop, 2007).

Limitations and future directions

This study presents a validated Persian version of the USDI for use with Iranian students. However, there are a few limitations that cannot go unmentioned. First, although the size of normative and clinical samples is adequate, the data were collected from only one university, and results may not be generalised to all Persian‐speaking students. Future studies should collect data from a number of Iranian universities. Second, the data are self‐reported and may be influenced by retrospective biases. Future studies should use interviews to check if the depression identified by the scale matches with the information obtained through other methods. Third, it is possible that the observed differences between males and females were superficial and confounded by other variables related to personality and/or socioeconomic factors. The way in which gender‐based behaviours developed in Western and Eastern cultures could impact patterns of depression should be investigated in the future through qualitative research. Future research must also assess the utility of this USDI adapted in Persian language as a screening and outcome measure in Iranian and other Persian‐speaking clinical populations. Fourth, measurement invariance techniques provide a far more robust test for gender differences across items. In the current study, sample sizes are not large enough to conduct these tests, but suitable caution over sum score of gender differences, and it suggested for future research.

Conclusion

The Persian version of USDI appears to be a promising tool to measure depression among Persian‐speaking university students in Iran and elsewhere. The scale is robust with a stable factor structure. Its psychometric properties are comparable with other widely used measures of depression. Although it is too early to draw conclusions about the USDI's clinical utility, the emerging evidence indicates its potential for identifying students at risk of depression in university counselling settings. Further, the scale can be useful for research purposes.

Declaration of Interest

The authors report no conflicts of interest. The authors alone are responsible for the content and authorship of this article.

Acknowledgement

We would like to offer our appreciation to the clinical staff and academics at the University of Tehran and its mental health centre for their support and assistance in data collection.