The Seasonal Pattern Assessment Questionnaire as a measure of mood seasonality: a prospective validation study

Abstract

The aim of this study was to test the validity of the Seasonal Pattern Assessment Questionnaire (SPAQ) as a measure of winter pattern seasonality of mood in the general population. The criterion was four repeated measurements of mood, collected in summer and winter each year for 2 years in a longitudinal questionnaire study. To account for the imperfect reliability of SPAQ reports, the SPAQ was completed on two occasions, and aggregate SPAQ variables created. Three variables from the SPAQ were tested for their correlation with prospectively measured lowering of mood in winter: seasonal pattern type, Global Seasonality Score (GSS) and self-reported problems with seasonal variation. The sampling frame was the electoral roll in a suburban region of Melbourne, Australia. Complete four-wave data were obtained from 304 respondents (53.9% female). Seasonal pattern type (self-report as feeling worst in winter months) was the only SPAQ variable to show a significant association with prospectively measured winter pattern seasonality of mood. It is provisionally concluded that the SPAQ can function as a measure of seasonality of mood in the normal population, but that validity has only been confirmed for the relatively gross variable of seasonal pattern type.

Introduction

Seasonal affective disorder (SAD) (Rosenthal et al., 1984) is typically understood as the extreme end of a continuum of normative seasonal variation in mood and behaviour (Hardin et al., 1991). In fact it has been suggested that the individual tendency towards seasonal variation in mood (‘seasonality’) may have more empirical support than categorical SAD (Bauer and Dunner, 1993). Seasonality is, therefore, a variable of interest in its own right, as demonstrated by its prominence in numerous studies (e.g. Eagles et al., 2002, Gordon et al., 1999, Hebert et al., 2002, Perry et al., 2001).

The Seasonal Pattern Assessment Questionnaire (SPAQ: Rosenthal et al., 1987) is a retrospective self-report instrument, originally designed to screen for SAD. Major strengths of the instrument are its brevity, high face validity and adequate psychometric properties (factor structure and test–retest reliability; see, Young et al., 2003). Therefore, it has been a foundation of research into both categorical SAD and the continuum of seasonality (see Mersch, 2001, for a summary).

There are four sections in the SPAQ: (1) degree of seasonal change in mood and behaviour (‘change items’); (2) pattern of seasonal change (‘calendar items’); (3) reactivity to different climatic and atmospheric conditions; and (4) whether (and, if so, to what extent) seasonal changes constitute a problem. Direction and extent of seasonality are measured separately on the SPAQ. The SPAQ's global seasonality score (GSS: the sum of responses to the six change items) is the most common operationalisation of the extent of seasonality. The direction of seasonality (‘seasonal pattern type’) has been typically measured on the response to 1 of the 10 calendar items of the SPAQ, namely, month(s) of feeling worst. Respondents are considered winter pattern if ‘feeling worst’ is associated with one of the winter months (and no summer months). Similarly, respondents are considered summer pattern if ‘feeling worst’ is associated with one of the summer months (and no winter months).

Remarkably, no published data have addressed the capacity of the SPAQ to predict individual differences in seasonality of mood in the normal population. A number of studies have tested the related issues of the SPAQ's validity as a screening instrument and its ability to distinguish between different SAD and non-SAD groups (see Mersch, 2001, for a summary). These studies have generally concluded that the SPAQ overestimates the prevalence of SAD (e.g. Levitt and Boyle, 1997), but adequately discriminates between SAD and non-SAD groups. As noted recently by Young et al. (2003) a fundamental question remains: is there a significant relationship between an individual's seasonality as measured on the SPAQ and as measured in actual mood variation across the seasons?

Longitudinal research across other time frames has consistently found that retrospective reports of mood variability are strongly influenced by preconceptions (Watson, 2000). For example, retrospective reports of feeling worst on Mondays are not supported by prospective data (Stone et al., 1985). It is, therefore, possible that retrospective reports of seasonal mood patterns also constitute popular misconceptions (Eastwood and Peter, 1988).

The present research sought to test the validity of the SPAQ as a measure of seasonality by comparing SPAQ-estimated mood seasonality against the criterion of prospectively measured seasonal mood variation. It is important to note here that the SAD syndrome is broader than seasonal variation in mood. The six change items on the SPAQ, for example, refer to mood, sleep, social activity, weight, appetite and energy. For the present study, however, it is assumed that changes in mood are a core feature of SAD and seasonality. From this perspective, the degree of concordance between SPAQ-estimated seasonality and prospectively measured seasonal mood variation constitutes a test of the SPAQ's validity.

Prospective mood seasonality was measured in winter and summer self-reports of current mood across 2 years in a community sample. On the basis of contemporary research into the structure and adaptive function of mood (Watson et al., 1999), behavioural engagement (BE) (Depue et al., 1987) was selected as the mood variable most likely to demonstrate seasonality in the general population.

Three self-report variables from the SPAQ were investigated as potential correlates of prospectively measured seasonality of mood: Seasonal Pattern Type (dichotomised into Winter and Non-winter types), problems associated with seasonal variation (Seasonal Problem, dichotomised into problems present or absent) and GSS. It was predicted that (i) participants who identified themselves as Winter type would exhibit more marked prospective winter pattern seasonality than participants who did not report this pattern, and (ii) GSS and Seasonal Problem would also correlate with the extent of prospective winter pattern seasonality, but only amongst self-identified Winter pattern respondents.