Participants
The study population was psychology students from a Peruvian public university located in Metropolitan Lima who were predominantly of lower middle or low socioeconomic level. All had Spanish as their first language. The nonrandomly accessible sample was students in their last two semesters of study, corresponding to the internship or preprofessional practice period in various Peruvian institutions. There were 125 students, with characteristics presented in Table 1, aged between 23 and 28 years, all born in Lima; the internship sites were distributed over a wide range of areas in Lima. The asymmetrical distribution of sex is typical in psychology careers in Peru. All were single, without family responsibilities, and eventually had formal part-time jobs. The exclusion criteria were not consenting to participate and having produced an infrequent response pattern (see Results section).
Instruments
Coping strategy indicator–Short version (CSI-S10)
The CSI-S is an abbreviated measure of coping strategies (problem solving, support seeking and avoidance; PS, SS, AVO, respectively), derived from the full Spanish version14, in which the respondent is asked to briefly write down a problematic situation important to him/herself and then answer 15 ordinally scaled items (not at all, a little and a lot; 5 items in each strategy). In the study of the adaptation, the reliability was satisfactory.
Single-item measure of stress (SIS1)
The SIS is a single-item measure that captures the experience of general stress based on emotional tension, restlessness and generalized worry, and was assessed with one item (Spanish translation: “Estrés significa una situación en que una persona se siente tensa, inquieta, nerviosa o ansiosa, que tiene problemas para dormir, debido a su mente está preocupada todo el tiempo. ¿Usted se siente así en estos días?”) scaled in five options (from ‘almost nothing’ to ‘a lot’). Due to its corroborated correlates with other variables, as an abbreviated measure, it represents an acceptable estimate of the overall stress experienced in several contexts17,18.
Efficacy to cope with difficulties (ECD)
The ECD item was developed as a proxy of the perceived global efficacy of coping with adversities experienced as difficulties. It was designed to represent the perceived effect on the ability to use coping strategies in the face of difficulties or stressors. It was phrased as “Puedo hacer frente a las dificultades (problemas) que se me presentan” (English translation: “I can cope with the difficulties (problems) that come my way"), scaled on five options (strongly disagree to strongly agree). This measure was influenced by the development of brief, efficient and valid measures applied in the work context19, and the global perception of effective coping linked to constructs measured with the CSI-S.
Procedure
Data collection
The collection was carried out between January and April 2019; it was face-to-face, in groups of between 3 and 10 participants. The order of the material was constant (informed consent, sociodemographic sheet and instruments). The entire procedure was performed according to the Declaration of Helsinki, emphasizing the voluntariness of participation, anonymity of response, confidentiality of data, and freedom to discontinue participation at any point. There were no material incentives for participation.
Analysis
First, a nonparametric framework was used to establish the scaling properties of the CSI-S items. The dimensionality was examined using the Poly-DETECT algorithm for ordinal items20, with confirmatory (to verify multidimensionality) and exploratory (to verify the number of dimensions) specifications. The following dimensionality indicators were used21: DETECT (> 1.0, strong multidimensionality), ASSI and RATIO (in both, 1.0 indicates dimensionally simple structure).
Second, the nonparametric item response theory model Mokken Scaling Analysis (MSA22) was used to a) corroborate scaling (coefficient H ≥ .3023), b) evaluate the model for monotonic homogeneity (cutoff for model acceptance, CRIT < 80), c) explore local independence (with coefficients, W(1), W(2) and W(3)24), and d) obtain a measure of reliability of the observed scores25. To reduce the impact of irrelevant or invalid responses for MSA, we used the G+ coefficient22, which identifies participants with scores that violate the Guttman model. The R program mokken22 was used.
Third, another nonparametric item response theory approach, Ramsay curves (RC26,27), was used to obtain optimal scores constructed by kernel smoothing in the regression between the items and the ranking of the individuals derived from the direct score. From this, regression weights were obtained to construct a maximum likelihood score (pML) for each subject, which maximized the validity of the scaling of subjects on the observed score15. The R program KernSmoothIRT28 was used.
In the analysis for assessing the relation with external variables, ordinal regression was applied, with the pML scores of the PS, SS and AVO scales as predictors of efficacy to cope with difficulties (ECD) and single-item stress (SIS), with a flexible threshold model. For the single-item stress response, the logit link function was used due to the low frequency of response in the highest response category (n = 4, 3.2%). To reduce convergence problems and large standard errors17, options 4 (n = 14, 11.2%) and 5 (n = 18, 14.4%) of the SIS items were combined with the adjacent category. For the ECD response, options 1 and 2 were combined, and the clog-log link function was used due to slight left skewness of their distribution. The R program ordinal29 was used.