Psychometric properties of the Inventory of Interpersonal Problems-64 in Colombia

Context and Objectives

The Inventory of Interpersonal Problems-64 (IIP-64) is one of the most widely used instruments to assess difficulties in interpersonal functioning. Derived from the original 127-item version (Horowitz et al., 1988), the IIP-64 represents a reduction to 64 items distributed across eight octant scales based on an octagonal circumplex model of interpersonal behavior (Alden, Wiggins & Pincus, 1990). This model is grounded in two primary dimensions: Affiliation (love) and Power (status, dominance). The IIP-64 has been studied in multiple clinical and research contexts, demonstrating sensitivity to change following therapeutic interventions.

The general objective of the present study was to analyze the reliability and validity of the Spanish-translated version of the IIP-64 in an incidental sample of 701 adult participants residing in Colombia. Specifically, the study aimed to examine the internal consistency of the scales, validate the circumplex structure using principal component analysis (PCA), and explore the correlations between the instrument and related measures of psychological distress, satisfaction with life, and obstacles to pursuing important life goals.

Method

Participants

A total of 729 participants were recruited through online advertisements and questionnaire response. After eliminating 28 records that did not meet inclusion criteria (being over 18 years of age and being Colombian or residing in Colombia), the final sample comprised 701 participants (387 women, 55.2%). There were two records with missing data (0.3%). Participants' ages ranged from 18 to 76 years, with a mean of 34 years (SD = 13.67). The most frequent place of residence was Bogotá DC (53.3%), with the remainder distributed across other cities in the country.

Instruments Evaluated

Inventory of Interpersonal Problems-64 (IIP-64). The IIP-64 is a self-report questionnaire consisting of 64 items presented on a 5-point Likert-type scale (0 = not at all, 4 = extremely). The instrument evaluates recent interpersonal difficulties that the person has experienced in their relationships with others, based on two main factors: Affiliation and Power. It contains eight subscales distributed in an orthogonal circumplex model with eight items each. In the original English construction, the IIP-64 obtained reliability coefficients ranging from .72 to .85 (Alden et al., 1990). The Spanish adaptation reported reliability coefficients ranging from .71 to .88 (Salazar et al., 2010).

Other Instruments

General Health Questionnaire-12 (GHQ-12). The GHQ-12 (Goldberg & Williams, 1988; Spanish version by Rocha et al., 2011) assesses emotional symptoms common in psychiatric disorders. It contains 12 items, with six expressed positively and six expressed negatively. The total score is an indicator of perceived psychological distress. The questionnaire obtained a reliability coefficient of .90 and demonstrated unidimensional structure (Ruiz et al., 2017).

Satisfaction with Life Scale (SWLS). The SWLS (Diener, 1984; Spanish version by Ruiz et al., 2019) consists of five items answered on a 7-point Likert-type scale, evaluating overall satisfaction with life according to the person's evaluation. Validation in Colombia conducted by Ruiz et al. (2019) demonstrated a reliability coefficient of .89 and unidimensional structure.

Valuing Questionnaire (VQ). The VQ (Smout, Davies, Burns & Christie, 2014; Spanish version by Ruiz et al., 2022) measures the consistency with which the person acts in accordance with their values during the past week. The instrument consists of 10 items in Likert format with seven response options. Validation in Colombia also demonstrated a two-factor structure with alphas of .83 and .82 for the Progress and Obstruction factors, respectively (Ruiz et al., 2022). The Obstruction factor was expected to produce moderate to high correlations with interpersonal distress, while the Progress factor was expected to correlate negatively with interpersonal distress.

Data Analysis

The translation of the IIP-64 was carried out from its original English version (Alden et al., 1990; Horowitz et al., 2003) by three bilingual psychologists with experience in cross-cultural adaptations. Three meetings were held to discuss differences and unify the translation of the instrument, seeking to achieve best translation practices adapted to the linguistic and semantic uses of Spanish in Colombia. The resulting unified version was the one used in the present research.

Regarding sampling, although the data collection strategy was characterized as "incidental" type (Otzen & Manterola, 2017), its extensive dissemination on social media platforms brought it closer to snowball sampling. The recommendation of having a sample of at least 300 participants was followed to conduct principal component analysis (Catena, Ramos & Trujillo, 2003).

Statistical analyses were performed using Jamovi 2.0 and JASP 0.16.1. Comparability was maintained with the two foundational studies of the IIP-64: the construction of the circumplex scales (Alden et al., 1990) and the corresponding national standardization (Horowitz et al., 2000).

The internal consistency of the scales was analyzed using Cronbach's alpha coefficient (α) and McDonald's omega coefficient (ω), with values above .20 considered as acceptable item-total discrimination indices (Díaz & Leyva, 2013; Ebel & Frisbie, 1986).

Principal component analysis (PCA) was performed based on eigenvalues equal to or greater than one, with Varimax rotation and using individualized scores from the eight octants (eight items each). After the procedure, PCA was performed again, this time without the three items affecting reliability (items 13, 28, and 44).

Results

Overall Reliability

Analysis of the overall reliability of the IIP-64 yielded the same value for both Cronbach's alpha (α) and McDonald's omega (ω) coefficients, both being .931. Overall reliability was categorized as excellent.

Subscale Reliability

Internal consistency coefficients for each subscale are presented in Table 1. Subscales with lower coefficients were Vindictive/Self-Centered (α = .691, ω = .675), Self-Sacrificing (α = .677, ω = .686), and Intrusive/Needy (α = .696, ω = .694), whereas subscales with the highest coefficients were Nonassertive (α = .785, ω = .795), Socially Inhibited (α = .793, ω = .791), and Cold/Distant (α = .815, ω = .825).

Items 44, 13, and 28 showed problematic item-total discrimination indices. Item 44 (belonging to the Domineering/Controlling subscale) showed a discrimination index of r = .01 (p = .797), a value well below the desired minimum of .20. After its elimination, α and ω values rose to .766. Item 13 (Nonassertive) showed an item-total discrimination index of r = .164 (p < .001). When item 13 was eliminated, coefficients rose to .814 for α and .817 for ω. Item 28 (Cold/Sacrificing) showed a low but significant discrimination index of r = .186 (p = .001). When item 28 was removed, the ω coefficient for the subscale was .701.

Overall reliability was reanalyzed without the three items affecting reliability (items 13, 28, and 44), obtaining α = .933 and ω = .932, with no increase in overall internal consistency (α = .933, ω = .932).

Construct Validity: Circumplex Structure

Pearson correlation coefficients were calculated between subscales using raw scores to verify the expected circular pattern in a circumplex model. The heat map presented in Figure 1 shows that the subscales met the condition p1>p2>p3>p4 expected in circumplex theory. The correlation with the adjacent scale, which is Vindictive/Self-Centered (r = .20), corresponds to p1; the correlation with the Cold/Distant subscale, which is two octants away (r = -.13), corresponds to p2; the correlation with the Socially Inhibited subscale which is three octants away (r = -.28), corresponds to p3; and, finally, p4 corresponds to the opposite scale, which is Nonassertive (r = -.47). Consistent with expectations, the r values indicate that as one moves further from the origin in circumplex space (leaving a distance of 0 as the origin, lower values indicate the condition .20>.13>.28>.47, supporting a pattern of circular correlations for this subscale).

Principal component analysis (PCA) performed on individualized scores of the eight octants (eight items each) yielded two factors: the first factor, labeled Affiliation, obtained an eigenvalue of 2.378 and explained 29.3% of the variance; the second factor, labeled Power, obtained an eigenvalue of 2.140 and explained 27.2% of the variance. These two factors explained 56.5% of the variance. The factors were effectively orthogonal with a correlation of r = 0.00. Plotting the interpersonal octants on a two-dimensional plane effectively derived the octagonal circumplex structure that theoretically supports the test (see Figure 2).

The same PCA was performed again, but this time without the three items affecting reliability. However, it was observed that the circumplex structure of the test was altered, distorting the distribution, angles, and vectors of the circumplex octants. Even when only item 44, which appears to be the least discriminant, was removed, the circumplex structure of the IIP-64 was distorted. For this reason, it was concluded that these three items are irrelevant to the construct validity of the test and that the test should be used in its entirety so that the slight sacrifice in the reliability of the respective scales would be justified by maintaining construct validity, and therefore it was concluded that it was better to retain these items.

Convergent Criterion Validity

Table 4 presents correlations of the IIP-64 and its subscales with related measures. Correlations of the IIP-64 with the GHQ-12 were positive and statistically significant. The strongest correlation was between the IIP-64 total score and GHQ-12 (r = .39). Correlations of the IIP-64 with SWLS were negative and statistically significant, indicating that greater interpersonal distress is associated with lower life satisfaction. The correlation between IIP-64 total score and satisfaction was inverse and of moderate size (r = -.317).

Correlations of the IIP-64 with VQ-Obstruction were positive and statistically significant, indicating that greater interpersonal distress is associated with greater difficulty moving in the direction of important life goals. The correlation of IIP-64 total score with VQ-Obstruction was moderate (r = .378). On the other hand, consistent with expectations, correlations of the IIP-64 with VQ-Progress were negative and statistically significant.

At a more detailed level, correlations between interpersonal problems and psychological distress were higher for Socially Inhibited, Overly Accommodating, Nonassertive, and Self-Centered. Furthermore, the interpersonal problem that showed the highest negative correlation with satisfaction was Socially Inhibited. Finally, the interpersonal problem that showed the highest correlations with VQ-Progress (negative) and VQ-Obstruction (positive) was Nonassertive.

Discussion and Conclusions

The general objective of the present study was to analyze the reliability and validity of the Spanish-translated version of the IIP-64 in an incidental sample of 701 adult participants in Colombia. For this purpose, an instrumental design was implemented and reliability was analyzed using Cronbach's alpha coefficient and McDonald's omega coefficient, and construct validity was explored by analyzing the circumplex structure using PCA and calculating correlations of the IIP-64 with other related measures.

Regarding instrument reliability, it is acceptable to good. Results of internal consistency of the scales estimated using Cronbach's alpha coefficients ranged from .68 to .82. Compared with these values and the foundational study of the IIP-64 (Alden et al., 1990), the lower end is one hundredth lower, and the upper end is three hundredths lower. These α coefficients are also slightly lower relative to findings by Holforth et al. (2006) and Salazar et al. (2010). However, the overall alpha coefficient of the IIP-64 in this study was slightly higher than that of Salazar et al. (2010). The overall performance of the IIP-64 as a measure of interpersonal distress with a total alpha coefficient of .93 can be categorized as excellent (George & Mallery, 2019).

Unlike previous studies, this research estimated test internal consistency using McDonald's omega coefficient with acceptable to good results, ranging from .68 to .83. The behavior of this reliability coefficient was very similar to that of Cronbach's alpha, providing additional convergent information on the internal consistency of the scales. Items 44, 13, and 28 appear to decrease the internal consistency of their respective scales. However, removing these items did not increase the overall internal consistency of the IIP-64 and affected the circumplex structure of the instrument; therefore, it was concluded that, in this sample, the complete set of items maintains the validity balance of reliability and validity of the set of items showing the best psychometric performance.

Regarding construct validity of the instrument, the behavior of the circumplex structure was observed to be comparable with the foundational study of the IIP-64 (Alden et al., 1990) and with the national standardization study in the United States (Horowitz et al., 2000). The IIP-64 evidenced reasonable approximation to an octagonal distribution in Cartesian space in terms of location, angles, and vectors of the scales. This distribution in circumplex space is superior to the dimensional scaling structure of the previous Spanish version of the IIP-64 (Salazar et al., 2010), thus overcoming possible reservations regarding construct validity.

Considering the convergence of construct validity and taking as reference instruments already validated in Colombia, correlations between constructs were also consistent with theoretical expectations. Taking the evidence as a whole, it can be concluded that the IIP-64 is a valid and reliable instrument for use in the Colombian population.

Some limitations of the present study merit mention. First, the psychometric properties of the IIP-64 were analyzed only in a general community sample. The absence of a clinical sample may have influenced the slightly lower values of internal consistency found. Second, correlations of the IIP-64 were obtained only with other self-reports, which could have inflated the obtained correlations. Third, the functioning of the IIP-64 has been tested only in Colombian participants, so we cannot generalize the results found in this study to other Spanish-speaking countries. Consequently, future studies could analyze the psychometric properties of the IIP-64 in other Spanish-speaking samples. Finally, the reliability of the IIP-64 was estimated only through the calculation of internal consistency coefficients, and future studies could analyze test-retest reliability of the IIP-64 and the sensitivity of the IIP-64 to interventions aimed at improving interpersonal functioning.

In summary, the IIP-64 demonstrated construct validity insofar as PCA yielded two main dimensions (Affiliation and Power) that exceeded the minimum criterion of explaining 50% of the variance (Merenda, 1997). Additionally, the scale complied with the expected octagonal circumplex structure, and its subscales showed a pattern of circular correlations. In addition, evidence of reliability using Cronbach's alpha coefficient and McDonald's omega coefficient was acceptable, exceeding coefficients of .60 and .70 (Loewenthal, 1996; Oviedo & Arias, 2005). Therefore, taking the evidence as a whole, it can be concluded that the IIP-64 is a valid and reliable instrument for use in the Colombian population.

Importance and Contribution

This study represents the first investigation reporting psychometric properties of the IIP-64 in Colombia and Latin America. Using a large community sample, acceptable to good psychometric properties were obtained for a Spanish version of the IIP-64, overcoming possible reservations about construct validity of previous Spanish versions.

The robust validation of the circumplex structure and evidence of reliability exceeding recommended criteria (.60 and .70) make the IIP-64 a recommended instrument for the assessment of interpersonal problems in clinical, research, and even personality screening contexts for personality disorders in the Colombian population. These satisfactory psychometric properties also allow its use for future research and for assessment of change following therapeutic interventions aimed at improving interpersonal functioning in Colombian clinical samples.

---