Prevalence of seasonal affective disorder in primary care; a comparison of the seasonal health questionnaire and the seasonal pattern assessment questionnaire

Abstract

Background: Prevalence rates of SAD suggested by previous studies have ranged from 1 to 12% depending on the diagnostic criteria used. The Seasonal Pattern Assessment Questionnaire (SPAQ), a widely used screening tool, has been shown to have low specificity for SAD. The Seasonal Health Questionnaire (SHQ) was designed to better reflect the clinical criteria for SAD and has been shown to have a higher specificity then the SPAQ in a psychiatric outpatient setting. Objective: The current study was designed to assess the validity of the SHQ in general practice against systematic research interviews, to compare the sensitivity, specificity and positive predictive values of the SHQ and the SPAQ and to use these data to estimate the prevalence of SAD in primary care. Methods: 809 subjects in a consecutive series of patients attending Southampton general practices in winter 2000/01 completed the SHQ followed by the SPAQ; 56 were interviewed using the Structured Clinical Interview for DSM (SCID). Results: The SHQ was more sensitive and specific than the SPAQ and had higher positive and negative predictive values in screening for SAD. The SPAQ indicated a prevalence of SAD of 10.7% (95% CI 8.6–13.1) while the SHQ provided a significantly lower estimate of 5.6% (95% CI 4.2–7.4).

Introduction

Seasonal affective disorder (SAD) is a condition where sufferers experience recurrent depressive episodes during the same season of the year (Rosenthal et al., 1984). The condition is likely to be the extreme end of the dimension of seasonality, i.e. the extent to which one is affected by the change in seasons (Thompson et al., 1988). Thus, for example in Montgomery County, USA, a telephone survey showed that 92% of the population reported some seasonal changes in mood (Kasper et al., 1989) making the definition of a threshold for clinical SAD a more than trivial task.

Although SAD is most typically characterised by the reversed vegetative shift of hypersomnia, hyperphagia and weight gain (Garvey et al., 1988), these symptoms do not appear in the DSM IV or ICD criteria, which require only seasonal recurrences of major depressive disorder, whether or not they have atypical symptoms. Thus there are two models of SAD, the seasonality dimension model, which requires only a score above threshold on a rating scale, and the diagnostic model, which requires clear cut episodes of mood disorder. The fact that atypical symptoms are common in nonseasonal depression and that assessing the periodicity of episodes is not straightforward means that separating SAD from nonseasonal recurrent depression is also a more than trivial task.

Nevertheless, the apparently selective response of SAD and sub-syndromal SAD to bright light therapy (Thompson, 2001, Partonen and Lönnqvist, 2000) indicates the need for diagnosis especially in primary care where most such patients will be found.

Perhaps partly for these reasons the prevalence of SAD remains unclear, with general population estimates ranging from 0.4 to 12% (Blazer et al., 1998, Dam et al., 1998). The gender ratio is also unclear, most studies showing an excess in women (Schlager et al., 1993) but one giving an excess in men (Blazer et al., 1998). Onset appears to be usually in the third decade (Rodin and Thompson, 1997) but it can rarely occur among children (Sourander et al., 1999) and the elderly (Michalak, 2000). There appears to be a relationship to latitude (Mersch et al., 1999), but this is not invariable as the prevalence in Iceland is surprisingly low (Magnússon and Stefánsson, 1993).

There are several reasons why prevalence studies may have given widely dispersed estimates. The first is that the illness can only be observed in one season and the prevalence may differ according to geographical location. Second, most studies suffer from recall bias as shown by the lower prevalence in prospective than retrospective studies (Nayyar and Cochrane, 1996). Thirdly, there are four different definitions of SAD in use, in addition to the threshold definition on the SPAQ; the original Rosenthal criteria (Rosenthal et al., 1984), DSM-III-R (APA, 1987), DSM-IV (APA, 1994) and ICD-10 (World Health Organisation, 1992). All of these are complex definitions, requiring a diagnosis of depression in each of the episodes and an assumption that all counted episodes were similar. A minimum number of the episodes must have occurred in the same season of the year (which must itself be defined). The Rosenthal, DSM IV and ICD-10 criteria only require two but the DSM III-R requires three episodes. The number of previous years during which the episodes took place is not defined in the criteria but often assumed to be the whole lifetime. The maximum number of allowed episodes of nonseasonal depression, and the ratio to seasonal episodes must be defined to demonstrate seasonality (Thompson and Cowan, 2001). The Rosenthal criteria allow no nonseasonal episodes, DSM IIR requires a ratio greater than 3:1, while DSM IV and ICD-10 leave it to clinical intuition.

These intrinsic difficulties in the epidemiology of SAD are further complicated by the differences in item content between the clinical criteria and the most commonly used screening tool (the SPAQ; Rosenthal et al., 1984) which asks about seasonal variation in sleep, socialising, mood, appetite, weight and energy level. A seasonality score of 11 and a global severity assessment of moderate is taken to indicate SAD (Kasper et al., 1989). This model, however, corresponds to the dimensional concept of SAD rather than the diagnostic one. The difference may explain the low specificity of the SPAQ (Raheja et al., 1996, Eagles et al., 1999) its poor discrimination between SAD, and other forms of depression, and its low positive predictive value (45–48%: Magnússon, 1996, Raheja et al., 1996). It also has poor test–retest reliability, and is a poor predictor of future seasonal episodes (Raheja et al., 1996).

We have previously described the development of the Seasonal Health Questionnaire (Thompson and Cowan, 2001) as an alternative approach to screening based on the diagnostic rather than the dimensional concept. It is a machine-readable form that has been structured to identify patients with SAD according to each of the four definitions. Unlike the SPAQ, it limits the time frame to the past decade. It begins by screening for depression and then explores the relationship between depressive episodes and the seasons. The structured of the questionnaire enables the patient to stop as soon as they fail to meet one of the obligatory criteria. Patients with a seasonal pattern who do not meet criteria for major depressive episode are categorised as sub-syndromal, as described in the DSM IV textual accompaniment to the criteria (p. 389).

In a previous study the SHQ and the SPAQ were compared in a small group of clinically diagnosed out patients with either SAD or nonseasonal major depressive disorder, and the SPAQ was shown to be a more sensitive but less specific screening tool (Thompson and Cowan, 2001).

• To assess the positive and negative predictive value of the SHQ and SPAQ against a structured research diagnostic interview (SCID).

• To estimate the prevalence of SAD in consecutive primary care attendees in Southampton.