Reliability and validity of a Portuguese version of the Young Mania Rating Scale

The reliability and validity of a Portuguese version of the Young Mania Rating Scale were evaluated. The original scale was translated into and adapted to Portuguese by the authors. Definitions of clinical manifestations, a semi-structured anchored interview and more explicit rating criteria were added to the scale. Fifty-five adult subjects, aged 18 to 60 years, with a diagnosis of Current Manic Episode according to DSM-III-R criteria were assessed using the Young Mania Rating Scale as well as the Brief Psychiatric Rating Scale in two sessions held at intervals from 7 to 10 days. Good reliability ratings were obtained, with intra-class correlation coefficient of 0.97 for total scores, and levels of agreement above 0.80 (P < 0.001) for all individual items. Internal consistency analysis resulted in an α = 0.67 for the scale as a whole, and an α = 0.72 for each standardized item (P < 0.001). For the concurrent validity, a correlation of 0.78 was obtained by the Pearson coefficient between the total scores of the Young Mania Rating Scale and Brief Psychiatric Rating Scale. The results are similar to those reported for the English version, indicating that the Portuguese version of the scale constitutes a reliable and valid instrument for the assessment of manic patients. Correspondence

In spite of this, the YMRS has some limitations, such as the absence of precise definitions of the items evaluated.Without operational definitions, each evaluator uses his/her previous experience with psychopathology as a reference point to define the item evaluated, impairing the reliability between evaluators.Furthermore, there is no systematic procedure for data collection with the YMRS.Structured interviews or interview guidelines may be particularly useful for individual evaluators and/or evaluators without extensive training in the recognition and conceptualization of psychopathology (14).
Although most of the controlled studies with manic patients use the YMRS, there are several studies that do not utilize a specific scale for mania or associate a standardized instrument with another one for the rating of severity of symptoms, such as the Brief Psychiatric Rating Scale (BPRS) (15) and the Clinical Global Impression (16).This suggests that mania rating scales are not as good as the scales developed for other psychiatric disorders (5,17,18) and highlights the need for studies of the psychometric qualities of available mania scales, as well as for the development of better instruments.
The importance of this work becomes clear when one considers that Bipolar Disorder is one of the most prevalent and disabling mental disorders (17), and new drug development requires sensitive and reliable instruments.In particular, a systematically validated instrument for the assessment of the severity of manic symptoms is not available in Portuguese-speaking countries.As a consequence, the purpose of the present study was to translate the English version of YMRS into Portuguese and to assess the reliability and validity of the Portuguese version.

Translation and adaptation of the Young Mania Rating Scale
Three researchers independently translated the YMRS into Portuguese from its original English version and the different versions were discussed and compared.A final version was obtained, back translated and compared to the original scale in English.The conclusion was that there were no relevant differences in the content of the two versions of the scale.Some modifications were made in order to adapt the instrument for systematic application based on operational definitions.A glossary of definitions for identifying the manifestations observed was introduced to improve precision in their observation and reporting.Each item of the scale was defined according to signs and symptoms described by Kaplan and Sadock (19) and presented in the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV) (20).
To facilitate and systematize the application of the rating scale, a semi-structured interview outline was formulated with guidelines for questions.These were based on questions used for the assessment of other instruments already standardized for the Portuguese language (21,22).The aim was to provide elements that would allow the assessment of the severity of each item of the scale at the end of the interview.
An introduction containing explicit comments on the methodology used in the interview and instructions on how to codify the scale were developed to provide information about the codification procedure.
Finally, changes were made in the description of operational definitions of the levels of severity to better clarify the baseline criteria for distinguishing one degree from the other.
The Portuguese version of the Young Mania Rating Scale was named "Escala de Avaliação de Mania" (EAM) and its final version is presented in Appendix 1.

Subjects
Fifty-five patients (31 females and 24 males) aged 18 to 60 years (mean ± SD: 32.2 ± 10.9) took part in the study.The diagnosis of Bipolar I Disorder, current manic episode, was confirmed using the Portuguese version (21) of SCID for DSM-III-R (23).Patients were hospitalized in the psychiatric unit of the University Hospital, Faculty of Medicine of Ribeirão Preto and/or in the Psychiatric Hospital Santa Tereza.
The experimental protocol was approved by the Hospital Ethics Committee and each patient plus one of his/her relatives signed an informed consent form.

Evaluators
Seven psychiatrists took turns in differ-ent rating pairs.Training started with joint reading of the instrument, followed by explanation and discussion about the recording protocol and the list of operational definitions of the manifestations to be observed.Then, for training purposes, interviews with non-subject manic patients were recorded on videotape.The group of evaluators watched the recorded interviews and each one rated the patients independently.For each interview, the scores of the individual items of the scale were compared and the differences between evaluators were identified and discussed.This procedure was repeated until the evaluators reached a minimal agreement of 70% on individual items of the scale in three consecutive and independent ratings.
Finally, individual ratings with the EAM were performed.They consisted of a live interview with a manic patient held by one of the evaluators and observed by one of the authors of this study, followed by calculation of the agreement index.

Data collection
During the initial 48 h of hospitalization, the patients who fulfilled the inclusion criteria were rated with the EAM and the BPRS -Bech's version (24) translated and validated into Portuguese (25).In this version of the BPRS, items are rated within 5 levels of severity (0 to 4).The interview for grading the EAM and the BPRS, which lasted 15 to 30 min, was performed by one of the evaluators and observed by another.Soon after, each participating evaluator independently completed the scales.Evaluators were instructed not to exchange information until the end of the study.

Statistical analysis
Data were analyzed statistically with the SPSS for Windows software, version 6 (SPSS Incorporation, 1989Incorporation, -1993)).The inter-rater reliability coefficient for individual and total scores of the EAM was calculated using the intra-class correlation coefficient (26), which is considered to be the most appropriate measure to assess reliability between evaluators (27).
Internal consistency was assessed by Cronbach's α.The Pearson correlation coefficient was employed to evaluate the concurrent validity of individual and total scores of the EAM and the BPRS.Wilcoxon's proof was used to assess the EAM sensitivity to alterations in the severity of symptoms in the same individual between the first and second rating (differences observed after 7 to 10 days).The Pearson moment-product correlation coefficient r was employed to determine the concurrent validity of the EAM with the BPRS.A correlation was taken to be significant at the level of P ≤ 0.05.

Results
Ninety-three paired ratings were analyzed to determine the EAM inter-rater reliability indices.They resulted in 186 completed protocol sets, 106 from the first evaluation and 80 from the second.
Table 1 shows the values of the intraclass correlation coefficient for the individual items and for the total score of the EAM.Levels of agreement above 0.80 (P < 0.001) were obtained for all the individual items and for the final scores in the two evaluations.In the second rating, the agreement indices were slightly better than the first rating for the total score as well as for individual items, except for items 1 (Elevated mood), 3 (Sexual interest), 6 (Speed and amount of speech), and 10 (Appearance).
Cronbach's α was calculated using analysis of variance applied to the eleven items of the scale for the 55 subjects for internal analysis of consistency.From the resulting co-variance matrix, an α = 0.67 was obtained for the scale as a whole and an α = 0.72 for each standardized item (P < 0.001).Most of the items correlated well with one another as well as with the total score, so that their removal would reduce the internal consistency of the instrument.However, this does not apply to three of the items -thought content, irritability and insight -which in this order were responsible for the progressive decrease of internal consistency.If only the thought content item were removed, Cronbach's α would be equal to 0.72 and α for standardized items would be equal to 0.74.If two items -thought content and irritability -were removed, there would be α = 0.73 for the scale and α = 0.74 for standardized items.Finally, if the three itemsthought content, irritability and insight -were removed, there would be α = 0.76 for the scale and α = 0.76 for the standardized items.
To determine the concurrent validity, the correlation between the individual items and the total score of the EAM with the individual items and the total score of BPRS applied to the 55 subjects was calculated.To this end, bivariate correlation measurements between the two instruments were made using Pearson's moment-product correlation coefficient r.All the individual EAM items presented significant correlation at the level of P < 0.05 with at least one BPRS item.Item number 6 -Speech (speed and amount) was the EAM item which correlated more often with most of the BPRS items (11 items), while item 3 (Sexual interest) and item 4 (Sleep) correlated with only three BPRS items.As to the correlation of the BPRS items, it was observed that item 2 -Psychic anxiety, item 13 -Psychomotor retardation and item 16 -Blunted or inappropriate affect -did not significantly correlate with any EAM item.
BPRS is a scale designed to evaluate psychotic patients, mostly schizophrenics.However, five of its items measure severity of symptoms frequently seen in manic patients.In the present study, these items presented a high correlation with six EAM items, as shown in Table 2.
All the individual EAM items presented a significant correlation (P ≤ 0.05) with the BPRS total score, except for item 4 (Sleep) and item 11 (Insight).The same happened with the BPRS individual items as related to the EAM total score, except for item 2 (Psychic anxiety), item 5 (Self-depreciation and guilt feelings), item 6 (Somatic anxiety), item 9 (Depressed mood), item 13 (Psychomotor retardation), and item 16 (Blunted or inappropriate affect).There was a high correlation between the two scales (r = 0.78; P = 0.0001) with respect to the total scores.

Discussion
The overall results obtained in the present study show better reliability levels than those obtained in the original study (5).This improvement is likely to be due to the adaptations of the rating protocol, such as adding the glossary of operational definitions describing the terminology of the scale more clearly and allowing more precise discrimination between the severity levels, making them more compatible with current concepts of mania.Moreover, the structured interview with guideline questions standardized the way information was obtained, minimizing the variability of interpretation among the different evaluators.Similar results have been reported by Williams with Hamilton's Depression Scale (28), by Lindström et al. (29) with the Positive and Negative Syndrome Scale and by Crippa et al. (22) with the BPRS.In the current study we could confirm that more structured instruments increase agreement levels, because even though most of the evaluators had little clinical experience, good reliability indices were achieved.
The validity indices achieved in the present study indicate that the EAM can measure severity levels of mania.The present values of the correlation indices between individual items and total scores are similar to those reported in previous studies on the psychometric qualities of mania rating instruments (1)(2)(3)(4)(5)8).
The lack of a mania rating instrument validated in Portuguese for standard comparisons limits the present analysis of the concurrent validity of the EAM.Nevertheless, good correlation indices between EAM and BPRS items were obtained, particularly those more specifically pertinent to mania symptoms.
From a theoretical point of view, five BPRS items related to the manic symptoms described by DSM-IV (20) and ICD-10 (30) were identified, presenting a significant correlation with the scale total score.Considering these five items as a BPRS subscale, a strong correlation was found between the EAM individual items and the total score, in contrast to the other BPRS items, indicating the EAM specificity for rating manic symptoms.The correlation coefficient between the EAM total score and the BPRS total score (r = 0.78) was similar to the correlation indices reported with the BMSRS (1), YMRS (5) and CARS-M (8).
In the present study, changes to the YMRS (5) were made in order to minimize the shortcomings of the original scale, such as the absence of operational definition of the evaluated items and of a systematic procedure to collect data.The results discussed above show that this aim was achieved without changing the rating procedure, damaging the structural coherence of the instrument or adding time for its administration.Therefore, the EAM structure specifically takes into account the range of behavioral manifestations of the manic episode, using a relatively streamlined rating procedure.
Since this sample had only patients with severe mania, it was not possible to determine whether the EAM applies to hypomania or to manic patients with either mild or moderate symptoms.In the same way, it was not possible to establish whether the instrument can distinguish mania from other psychiatric disorders, particularly schizophrenia and other psychotic disorders.
Because the study was conducted without specifying the type of treatment or its duration, it became difficult to assess systematically the evolution of symptoms or to proceed to the longitudinal follow-up of the patient-subjects for a longer period of time.Thus, it was not possible to establish a cutoff EAM criterion predictive of either discharge or change in treatment modality.
EAM is the first instrument intended specifically for manic patients that has been adapted to the Portuguese language and subsequently validated by empirically studying its psychometric properties.The instrument shows high reliability and validity indices comparable to those previously reported with the use of similar instruments in the English language (1)(2)(3)(4)(5)(6)(7)(8)10,13,(31)(32)(33).The EAM is an instrument that can be easily applied and requires little training for its correct use.The use of a semistructured interview with guideline questions together with detailed operational definitions seems to have improved the instrument's reliability.As a result, the EAM may be used for assessing the severity of manic symptoms as well as to measure the efficacy of psychological and/or pharmacological treatment.