Testing the PROMIS^® Depression measures for monitoring depression in a clinical sample outside the US

Highlights

• PROMIS Depression shows excellent measurement properties in a clinical sample from Spain.

• Scores detect depression with similar ability than other common depression measures.

• Scores discriminates between major depression and comorbid anxiety disorders.

• Results support the usefulness of PROMIS Depression for cross-national comparisons.

• PROMIS depression can be used for monitoring depression in clinical settings.

Abstract

The Patient Reported Outcomes Measurement Information System (PROMIS) was devised to facilitate assessment of patient self-reported health status, taking advantage of Item Response Theory. We aimed to assess measurement properties of the PROMIS Depression item bank and an 8-item static short form in a Spanish clinical sample. A three-month follow-up study of patients with active mood/anxiety symptoms (n = 218) was carried out. We assessed model unidimensionality (Confirmatory Item Factor Analysis), reliability (internal consistency and Item Information Curves), and validity (convergent-discriminant with correlations; known-groups with comparison of means and effect sizes; and criterion validity with Receiver operating Characteristics (ROC) analysis). We also assessed 3-month responsiveness to change (Cohen's effect sizes (d) in stable and recovered patients). The unidimensional model showed adequate fit (CFI = 0.97, RMSEA = 0.08). Information Curves had reliabilities over 0.90 throughout most of the score continuum. As expected, we observed high correlations with external self-reported depression, and moderate with self-reported anxiety and clinical measures. The item bank showed an increasing severity gradient from no disorder (mean = 48, SE = 0.6) to depression with comorbid anxiety (mean = 55.8, SE = 0.4). PROMIS detected depression disorder with great accuracy according to the area under the curve (AUC = 0.89). Both formats, item bank and short form, were highly responsive to change in recovered patients (d > 0.7) and had small changes in stable patients (d < 0.2). The good metric properties of the Spanish PROMIS Depression measures provide further evidence of their adequacy for monitoring depression levels of patients in clinical settings. This double check of quality (within countries and populations) supports the ability of PROMIS measures for guaranteeing fair comparisons across languages and countries in specific clinical populations.

Introduction

Certain areas of medicine have a sustained interest in the development of Patient Reported Outcome (PRO) instruments (Black and Jenkinson, 2009). This interest has been accompanied by a proliferation of condition-specific instruments, causing a fragmentation of measures that hampers comparability across studies, settings, or pathologies. As a response, the Patient-Reported Outcomes Measurement Information System (PROMIS^®) (Cella et al., 2007) was devised in the US as a publicly available measurement system of self-reported health based on a domain-specific approach without attributions to specific conditions or treatments (Cella et al., 2010). PROMIS focuses on comparability between health states and populations through the application of item response theory (IRT), a psychometric method for item-calibration allowing a common metric for different populations, broader range of scores, and greater precision in individual measures compared to classical test theory methods. IRT properties yield the possibility of alternative administration forms: full item banks, static short forms or dynamic computer adaptive testing (CAT) that selects items in real time targeted to the examinee's specific level of ability or impairment, reducing the number of questions needed and respondent burden without a substantial loss of precision (Hambleton et al., 1991, Van der Linden and Glas, 2000). However, administration burden is increased as CAT requires computerized support in applications.

The international extension of PROMIS is currently underway (Alonso et al., 2013) with PROMIS domains being culturally adapted into several languages (Patient-reported outcomes measurement information system, 2015a). To support the usefulness of PROMIS^® for cross-national comparisons, it is important to demonstrate that PROMIS measures are valid, reliable and responsive to change when used outside the US. The assessment of cross-cultural differential functioning at the item (DIF) and test (DTF) level is also crucial to ensure that items are similarly understood and the measures are unbiased across different subpopulations, most importantly, countries, cultures and conditions.

A case of particular importance is emotional disturbance and depression, constructs negatively influencing the course of health (Anderson et al., 2001, Scott et al., 2009) that have been recommended as main outcomes to assess the impact of treatments for various specific conditions (Turk et al., 2003). Efforts have been made to develop item banks for CAT depression instruments (Fliege et al., 2005, Forkmann et al., 2013, Gardner et al., 2004, Gibbons et al., 2008, Gibbons et al., 2012). Among them, the PROMIS system includes a depression domain as part of the overall health profile; it is also the only IRT-based depression measure available in Spanish. An interesting feature of PROMIS Depression is that it does not include items regarding somatic symptoms (e.g. sleep problems, appetite disturbances), unlike other commonly used depression measures (Beck et al., 1996, Spitzer et al., 1999). Thus PROMIS avoids potential confounding effects when assessing patients with comorbid physical conditions. Another advantage of PROMIS measures is that they are designed to be population-independent and sensitive to prevalence but also to a wide range of severity levels. The dimensional approach also allows averting difficulties related to changes in the consensus criteria of categorical nosologies, a problem which is known to have a great impact in clinical patient status when it comes to modification of disorder compulsory criteria (Pereda and Forero, 2012). Additionally, it can provide valuable information on real or biased cross-national differences in the epidemiology of depression (Forero et al., 2014b) (Weissman et al., 1996).

In order to gain evidence about their usefulness, PROMIS Depression attributes should be tested in clinical environments in different languages. Of greatest concern is the evaluation of construct validity and responsiveness in patient samples relevant to the construct of interest. PROMIS Depression has shown good results in patients with major depression (Pilkonis et al., 2014) and other conditions (Amtmann et al., 2014). However, the psychometric properties of the PROMIS Depression measures in Spanish or other language versions have not been evaluated so far.

This study aimed at testing the measurement properties of the Spanish version of PROMIS^® Depression in patients seeking mental health care at different care levels in Spain. Specifically, our objectives were to: a) confirm the measurement model and unidimensionality of the PROMIS Depression item bank; b) assess reliability, construct-related validity and responsiveness to change of the item bank and the 8-item static short form.