|
Epidemiological surveys over the past decade have suggested that
the prevalence of the seasonal subtype of major depression (SAD)
is affected by various factors including age(1,2), sex (3), and
possibly latitude (2,4–6). However, these studies evaluated
the prevalence rates in different cities or regions that happened
to be at different latitudes and did not specifically use latitude
in setting up their sample. Other studies have found no significant
differences in the prevalence of SAD across large geographic regions.
Using data from the National Comorbidity Survey, Blazer and others
divided their sample geographically into the 4 quadrants of the
US (that is, the northwest, northeast, southwest, and southeast)
and found no difference in odds ratio (OR) for the presence of SAD
in these regions (7). Similarly, Saarijarvi and others compared
northern and southern Finland and found no significant differences
in prevalence of SAD (8). However, the division of countries into
such large areas (such as north and south) may not detect smaller
or subtler latitudinal changes. We designed this study to determine
the prevalence of SAD across the province of Ontario by dividing
the province into equal bands of latitude and sampling equally across
latitudes.
Methods
This study involved a highly structured telephone survey that used
a modified version of the Depression and Seasonality Interview (DSI).
The DSI was developed to identify subjects who met DSM-IV criteria
for major depression and mania and to evaluate the seasonal patterns
of these conditions. A new instrument was required because at that
time no existing structured interview made the diagnosis of SAD.
The DSI obtains the following information: 1) household demographics,
which allows both assessment of eligibility and description of the
sample; 2) all DSM-IV criteria for a current or past episode of
major depression and mania; 3) the burden of these episodes with
respect to hospitalization, medications, and dysfunction; 4) whether
there is a medical or physical condition or drug use that could
substantially explain all the symptoms of depression; and 5) the
seasonal variation in all 9 symptoms of depression. The current
modification, the DSI-Ontario Version (DSI-O), evaluates the seasonal
pattern of both depression and mania using an adaptation of the
Composite International Diagnostic Interview (CIDI) seasonal pattern
addendum (Kessler 1997, personal communication). It permits either
the worst, the first, the longest, or the most recent episode of
depression or mania to be selected as the index episode. In our
study, if one of these episodes did not reach criteria for major
depression or mania according to DSM-IV, or if the episode was judged
to be entirely due to medical illness, medications, drugs or alcohol,
or grief, all other potential episodes of depression or mania were
probed sequentially until either an episode of major depression
or mania was identified or the subject had no further episodes to
discuss.
Sampling Modelling Research Technologies, a marketing company based
in Toronto, ascertained telephone numbers from the Teledirect Directory,
which supplies telephone numbers with the postal code of household
dwellings in Ontario. Exact latitude and longitude can be determined
from these postal codes. For sampling purposes, we divided the province
into 8 equal strata between 41.50 °N and 49.49 °N. Population
density is extremely low above that range in latitude, and the cost
of sampling is prohibitive. Respondents lived in “census subdivisions,”
a term that refers to municipalities, as determined by provincial
legislation (9). For this study, each census subdivision was categorized
as rural if the centre had a population of less than 30 000 and
urban if the population was 30 000 or more. We made an a priori
rule that no stratum in our sample would be less than 18% rural
or 18% urban. We did this to ensure that numbers of respondents
would be sufficient to make meaningful comments in strata dominated
either by urban populations (such as stratum 3, which contains the
capital, Toronto) or rural populations (for example, stratum 8).
Equal numbers of telephone numbers for each stratum were randomly
selected and then randomly assigned for contact in each of the 4
seasons of the year (based on the equinoxes and solstices). Previous
investigations have demonstrated an effect of interview time of
year on the prevalence estimates of SAD (1). If there was a language
difficulty with the initial respondent (the person answering the
phone), the call was terminated and a substitute number contacted.
If the initial respondent did understand English or French, the
household’s eligibility was assessed with a brief introductory
screening questionnaire.
|
|
This questionnaire ascertains the age and sex of all occupants
aged 20 years or older. Households were eligible if at least 1 member
of the household was aged 20 years or older (and might therefore
have had at least 2 annual adult major depressive episodes) and
had lived more than 6 months each year, for at least 3 years, at
the current abode or within 150 kilometres (to ensure sufficient
and stable exposure to the risk factor, latitude). From the list
of eligible household members, the person with the next birthday
was then identified as the target respondent. If a target respondent
refused to participate, a substitute phone number was contacted.
Goldfarb Consultants, a market research company involved in the
development of the DSI and experienced in its use, conducted all
interviews.
The data collection methods and the interview used in this study
were otherwise similar to those in the previous Toronto survey (1),
except that the interview was translated into French to ensure that
Franco-Ontarians were also recruited. The diagnosis of SAD was made
according to DSM-IV criteria (that is, seasonal course specifier).
To score positive on the criterion “seasonal episodes substantially
outnumber non-seasonal episodes,” subjects had report that
at least 66% of all of major depressive episodes followed the typical
seasonal pattern (7). The DSM-IV requires the last 2 episodes to
be seasonal. We considered this definition to represent “current”
SAD. If subjects met all other criteria for SAD and had 2 episodes
in consecutive years at any time, we considered them to have “lifetime”
SAD. Therefore, a subject with current SAD, by definition, also
had lifetime SAD, but the reverse is not necessarily true. The DSI
also contains items from the Seasonal Pattern Assessment Questionnaire
(SPAQ) (9), a widely used screening scale that evaluates the lifetime
pattern of seasonal variation of 6 symptoms of depression. The total
score for these 6 items is known as the Global Severity of Seasonality
(GSS) score. The SPAQ also evaluates the degree to which these symptoms
as a whole interfere with functioning. An alternate diagnosis for
SAD was made using the SPAQ (SPAQ-SAD) according to the criteria
of Kasper and others (3).
During the course of the study, randomly selected individuals were
reinterviewed by telephone within 1 month of their original interview.
The entire interview was repeated on a second occasion, and the
subjects were provided $5 for their time. Because we did not see
the individuals face to face, we only accepted data from subjects
whose demographic variables matched exactly in both interviews.
Prevalence rates are presented descriptively. The difference in
prevalence across the 8 strata for SAD, SPAQ-SAD, and major depression
was evaluated using chi-square and chi-square for linear associations.
Difference in prevalence of SAD between men and women and across
the decades (that is, age 20 to 29 years, 30 to 39 years, and so
on) was tested using chi-square. The bivariate relation between
the GSS and latitude was tested using Pearson product-moment correlation
coefficients. For the reliability analysis, the kappa statistic
was used for test-retest reliability for the presence of major depression.
Results
Sample Description
Between mid-March 1996 and mid-March 1997, 6666 telephone numbers
were processed. No contact was made with 2191 (that is, there was
no answer, a business line, answering machine, or fax line was reached
, or there was a language difficulty). Of the 4475 successful contacts,
2397 refused outright to participate and did not hear any details
of the study. Of the remaining 2078, who heard details of the study,
228 (11%) refused to participate; 145 (7%) had no eligible member
of the household; 100 (5%) did not have complete data, or had language
difficulties, or terminated during the interview; and 1605 (77%)
completed the interview. Approximately 5 months into the study,
we noted a low recruitment in strata 6 and 7. We therefore increased
sampling in these strata for 3 months. This effort was very successful
and accounts for disproportionately high recruitment in these 2
strata during the fall months.
|
