Study procedure

The Depression Scale includes ten items, with four response alternatives scoring 0–3: ‘not at all’, ‘a little’, ‘quite a lot’ and ‘extremely’ (see Table 2). In the baseline study the cut-off point for the screening sum score was >8.

Table 1 Sensitivity and specificity of different Depression Scale cut-off points

Depression scale score	Sensitivity (%)	Specificity (%)
Score at follow-up v. CIDI–SF
8	95.2	74.4
9	95.2	78.5
10	90.5	83.6
11	90.5	86.8
12	81.0	89.5
13	76.2	91.3
14	71.4	92.2
15	71.4	94.1
Score at baseline v. CIDI–SF
8	95.5	27.3
9	90.9	41.2
10	86.4	53.7
11	86.4	62.5
12	72.7	71.8
13	72.7	77.8
14	63.6	83.8
15	59.1	87.5

CIDI–SF, Composite International Diagnostic Interval–Short Form

Table 2 Sensitivity and specificity of Depression Scale items at baseline and at follow-up compared with depression assessment with the Composite International Diagnostic Interview.

	DEPS score v. CIDI–SF episode of depression
	DEPS score at follow-up			DEPS score at baseline
Depression Scale items 1	Sensitivity (%)	Specificity (%)	Sensitivity (%)	Specificity (%)
During the past month I have…
1… suffered from insomnia	63.6	84.4	54.5	80.0
2… felt blue	59.1	89.3	72.7	74.8
3… felt everything was an effort	81.8	86.4	86.4	63.3
4… felt low energy or slowed down	72.7	83.8	59.1	66.4
5… felt lonely	59.1	93.9	22.7	81.9
6… felt hopeless about the future	81.8	92.5	59.1	74.7
7… not got any fun out of life	54.5	84.8	27.3	73.5
8… had feelings of worthlessness	50.0	96.1	45.5	81.0
9… felt all pleasure and joy has gone from life	45.5	93.8	59.1	82.8
10… felt that I cannot shake off the blues even with help from family and friends	42.9	91.2	36.4	81.0

CIDI–SF, Composite International Diagnostic Interview–Short Form; DEPS, Depression Scale

1. All items are scored 0, not at all; 1, a little; 2, quite a lot; 3, extremely. An item was regarded as positive when the score was >1

In the follow-up study participants again filled in the Depression Scale, the Michigan Alcoholism Screening Test (Reference SelzerSelzer, 1971), parts of the Hopkins Symptom Checklist (Reference Derogatis, Lipman and RickelsDerogatis et al, 1974), and structured questions. To assess major depressive episode, 38 items from the Short Form of the Composite International Diagnostic Interview (CIDI–SF; World Health Organization, 1989; Reference Kessler, Andrews and MroczekKessler et al, 1998) were used in a telephone interview. The CIDI–SF questions concerning the occurrence of symptoms of a major depressive episode referred to the previous month. Three trained psychiatrists (A.M. and Drs Liisa Groth and Niko Seppälä), each with at least-years' experience in psychiatry, conducted the interviews, masked to the baseline PSE diagnoses.

Statistical methods

The accuracy of the Depression Scale as a screening instrument for depression was assessed by receiver operating characteristic (ROC) curve analyses. The follow-up Depression Scale score (DEPS–F) was compared with the CIDI–SF diagnosis of depression. The ability of the baseline Depression Scale score (DEPS–B) to predict the CIDI–SF diagnosis at follow-up was also evaluated. In ROC analyses, sensitivity, specificity and areas under the curve were calculated. Sensitivity and specificity were calculated for each reasonable cut-off point of the Depression Scale.

To evaluate which single items of the DEPS–B were best at predicting a depressive episode, the sensitivity and specificity for single items were calculated. After that, logistic regression analysis with forward stepwise method using all DEPS–B items as predictors was conducted. For this analysis, all items were dichotomised using 1 as the cut-off score (0–1, negative item result; 2–3, positive item result). To evaluate which single items of the DEPS–F were best for recognising a depressive episode, the sensitivity and specificity were calculated separately for each item, and logistic regression analysis was likewise conducted.

To identify an ideal pair of Depression Scale items for composing a short version of both DEPS–B and DEPS–F, sensitivity and specificity for every possible DEPS–B and DEPS–F item pair were calculated. An ideal pair of items implied that both of the items scored above 1. Only pairs in which sensitivity was at least 50% were regarded as relevant and reported.

Analyses were performed using the Statistical Package for the Social Sciences version 11.5 for Windows; P<0.05 was considered statistically significant.

Depression Scale v. CIDI-SF

In participants with CIDI–SF depression, the median DEPS–F score was 18 (range 7–30) and in those without depression it was 5 (range 0–28) (P<0.001, Mann–Whitney test). In the ROC analysis of DEPS–F v. CIDI–SF the area under the curve was 0.939 (Fig. 1). The ideal pair of sensitivity (90.5%, 95% CI 0.71–0.97) and specificity (86.8%, 95% CI 0.82–0.91) was found with a score of >11 as the cut-off point (Table 1). In participants with CIDI–SF depression the median DEPS–B score was 17 (range 2–24) and in those without depression it was 10 (range 0–28) (P<0.001, Mann–Whitney test). In the ROC analysis of DEPS–B v. CIDI–SF the area under the curve was 0.803 (Fig. 1). The ideal pair of sensitivity (72.7%, 95% CI 0.52–0.87) and specificity (77.8%, 95% CI 0.72–0.83) was found with a score of >13 as the cut-off point (Table 1).

Fig. 1 Receiver operating characteristic curves: (a) Depression Scale score at follow-up v. Composite International Diagnostic Interview-Short Form (CIDI-SF) depression at follow-up; (b) Depression Scale score at baseline v. CIDI-SF depression at follow-up.

Depression Scale items v. CIDI-SF

The three most sensitive DEPS–F items were 3 (‘I have felt everything was an effort’), 6 (‘I have felt hopeless about the future’) and 4 (‘I have felt low energy or slowed down’), and the most specific items were 8 (‘I have had feelings of worthlessness’), 5 (‘I have felt lonely’) and 9 (‘I have felt all pleasure and joy has gone from life’) (Table 2). In the case of DEPS–B, item 3 had a high sensitivity whereas items 9, 5, 8 and 10 (‘I felt that I cannot shake off the blues even with help from family and friends’) had a reasonably high specificity. One item (item 3) was quite sensitive in both analyses, for both recognising and predicting CIDI–SF depression.

In logistic regression analyses, DEPS–F items 3 and 6 were significantly associated with CIDI–SF depression, whereas DEPS–B items 1 (‘I have suffered from insomnia’), 3 and 9 significantly predicted occurrence of subsequent CIDI–SF depression (Table 3).

Table 3 Depression Scale items at baseline and at follow-up from logistic regression analyses significantly associated with depression at follow-up assessment.

Depression Scale items	OR	(95 % CI)	P
DEPS at follow-up v. depression at follow-up (CIDI–SF)
During the past month I have…
3…felt everything was an effort	5.54	(1.35–22.79)	0.017
6…felt hopeless about the future	21.89	(5.45–88.01)	<0.001
DEPS at baseline v. depression at follow-up (CIDI–SF)
During the past month I have…
1…suffered from insomnia	2.67	(0.99–7.19)	0.055
3…felt everything was an effort	6.50	(1.76–24.01)	0.001
9…felt all pleasure and joy has gone from life	3.70	(1.35–10.09)	0.011

CIDI–SF, Composite International Diagnostic Interview–Short Form; DEPS, Depression Scale

Best Depression Scale item pairs v. CIDI-SF

Sensitivity and specificity were calculated for every possible pair of Depression Scale items to ascertain which two items had the best balance of recognition and prediction. Only the pairs with sensitivity of at least 50% are reported (Table 4). The three best pairs for recognition were items 3 and 6, items 3 and 4, and items 4 and 6, whereas the best pairs for prediction were items 2 (‘I have felt blue’) and 3, items 3 and 4, and items 3 and 9.

Table 4 Sensitivity and specificity of Depression Scale item pairs at baseline and at follow-up compared with depression at follow-up assessment

	DEPS score v. CIDI–SF episode of depression
	DEPS at follow-up			DEPS at baseline
DEPS item pair 1	Sensitivity (%)	Specificity (%)	Sensitivity (%)	Specificity (%)
1+3	59.1	93.3	45.5	89.3
1+6	54.5	96.4	31.8	93.7
2+3	50.0	93.3	72.7	84.0
2+6	59.1	96.4	50.0	87.5
2+9	27.3	96.0	54.5	91.2
3+4	72.7	89.0	59.1	76.9
3+5	54.5	96.5	22.7	89.3
3+6	72.7	95.6	54.5	84.9
3+9	45.5	95.6	59.1	88.9
4+6	68.2	95.2	36.4	88.4
5+6	54.5	96.9	13.6	89.7

CIDI–SF, Composite International Diagnostic Interview–Short Form; DEPS, Depression Scale

1. Item pair is included in the table when sensitivity was >50.0% in either of the analyses. A DEPS item was regarded as positive when the score was >1

Sensitivity and specificity

The first validation of the Depression Scale was reported in an earlier study, in which the cut-off point for depression was >8 (Reference Salokangas, Poutanen and StengardSalokangas et al, 1995). In the baseline validation study, using the PSE as the criterion, the sensitivity of the Depression Scale for clinical depression was 74% and the specificity for non-depression 85%. For severe depression the figures were 84% and 93%. In the present study the figures for sensitivity and specificity were better than those of the earlier validation study. In the baseline validating analyses the sampling ratio was taken into account, but this was not done in the present study, which was mainly intended to ascertain the ability of the scale to predict an episode of depression and to evaluate its individual items. The differences in the levels of sensitivity and specificity between the baseline validation analyses and these follow-up analyses are perhaps partly explained by this fact. There are also differences in the validity criterion between the two diagnostic instruments. The PSE is based on symptoms, and the CIDI is based on syndromes (Reference Lowe, Spitzer and GrafeLowe et al, 2004). With the CIDI–SF the definition of depression was clearer because there were only two categories: depressive and non-depressive. It should also be kept in mind that the PSE interviews at baseline were held face-to-face, whereas the CIDI–SF interviews at follow-up were conducted by telephone. A telephone interview relies more on the examinee's own assessment, and is closer to a self-rating instrument like the Depression Scale. The same items of the CIDI–SF were used as in a previous Finnish depression study (Reference Isometsa, Aro and AroIsometsa et al, 1997; Reference Lindeman, Hamalainen and IsometsaLindeman et al, 2000) using the computer-assisted telephone interview method.

According to Lowe et al (Reference Lowe, Spitzer and Grafe2004) the sensitivity of screening questionnaires should lie above specificity and be as high as possible, and the specificity should be at least 75%. In this study the cut-off point >11, which has a sensitivity of 90.5% and specificity of 86.8%, could be ideal.

When the ability of the Depression Scale to predict an episode of depression was analysed, the area under the curve was 0.803. An earlier study with primary care patients (Reference Salokangas, Stengard and PoutanenSalokangas et al, 1994) showed that the rate of clinical depression in people with a Depression Scale score above 12 was about 47% and in those with a score above 15 it was about 57%. These percentages are high enough to have some clinical value. In this study, with a cut-off point of >11 sensitivity was 86.4% but specificity only 62.5%. When an instrument is used as a predictor it is perhaps more important to avoid false positives and not to stigmatise patients; this justifies a higher cut-off point.

What did the Depression Scale actually assess?

In a study of general practice patients (Reference Williamson, Neale and SterneWilliamson et al, 2005), four mental health self-report scales and a composite of those four were assessed to determine their accuracy in predicting psychiatric caseness for depression, dysthymia, generalised anxiety disorder, social phobia, agoraphobia and panic attack. One scale measuring neuroticism – the Neuroticism Scale of the Eysenck Personality Questionnaire (EPQ–N; Reference Eysenck, Eysenck and BarrettEysenck et al, 1985) – and a composite of all four scales were found to be very strong and accurate predictors of psychiatric caseness, but they were unable to differentiate between specific disorders. In our study only episode of depression – not other psychiatric diagnoses – was assessed.

In an extensive follow-up study (Reference Tyrer, Seivewright and JohnsonTyrer et al, 2004) the quick-to-use HADS was good for recognising both depression and anxiety, and was better than any other single measure for predicting the outcome of both anxiety and depressive disorders after an interval of 12 years. The Montgomery–Åsberg Depression Rating Scale did not have such predictability.

When the Depression Scale and two common self-rating instruments (the BDI and the HADS) are compared, they differ in many ways. The Depression Scale concentrates on the previous month, whereas the BDI concentrates on the previous week (the BDI–II on the past 2 weeks; Reference Beck, Steer and BallBeck et al, 1996) and the HADS on current feelings. Of the criterion standards used in this study, both the PSE and the CIDI–SF refer to the previous month. It is difficult to say, however, what the true significance of the differences in these time periods is.

The Depression Scale is the shortest of the three instruments, and the BDI is the longest. The formulation of the items is different: the most evident difference is that the Depression Scale gives exactly the same short-answer alternatives for all ten items, whereas there are several different sets of alternative answers in both the BDI and the HADS. This makes the Depression Scale very quick and easy to use, and increases adherence.

The BDI includes most of the Depression Scale topics. Only the topics of items 5 (loneliness), 7 (no fun) and 10 (not helped even with family and friends) are missing in the BDI. The Depression Scale item 5 was specific in recognising depression and item 10 specific in predicting it. However, the BDI covers the symptoms of depression more comprehensively than the former scale. The HADS covers both depression and anxiety, but lacks most of the Depression Scale topics (items 1, 2, 3, 5, 8 and 10); the symptoms covered are less severe than in the BDI or in the Depression Scale. Common topics for all the three self-rating instruments are the Depression Scale items 4 (low energy), 6 (hopelessness), and 9 (lost pleasure and joy). These topics probably relate to the core of depression symptomatology; other topics can be said to be consequences of the core symptoms and not so essential to depression only.

The Depression Scale items 3 and 4 were good at both recognising and predicting depression. Item 3 (‘I have felt everything was an effort’) suggests reduction of energy, which is one of the main symptoms of depression according to the ICD–10. Item 6 was good for recognition even though its wording refers to the future (‘I have felt hopeless about the future’); hopelessness is also a symptom of depression in the ICD–10. Item 9 was good in predicting depression. The wording of item 9 (‘I have felt all pleasure and joy has gone from life’) refers to something that has already happened, something that is possibly endured as beyond help. Item pair 2 and 3 was the best at predicting depression. The wording of item 2 (‘I have felt blue’) may be experienced as persistent low mood, referring to a more chronic state. It is almost the same as lowering of mood, one of the main symptoms of depression in ICD–10. The best combination – and a possible quick version – of two items for recognising depression was items 3 and 6, and the best combination for predicting depression was items 2 and 3.

The use of psychometric scales is in general problematic. Among people who appear to be healthy according to standard mental health scales it is possible to identify a subgroup of people who may not be psychologically healthy at all: mental health scales may assess not mental health but instead defensive denial (Reference Shedler, Mayman and ManisShedler et al, 1993). Moreover, any scale that is valid for assessing current depression will have some long-term predictability because depression is recurrent. However, if a scale has predictability, it means it has the ability to catch not just reactive and short-term symptoms but more chronic or recurrent core features of the disorder.

Limitations and strengths of the study

It is a limitation of the study that the interviews were held by telephone. However, the CIDI–SF telephone interviews were conducted with care and by experienced psychiatrists. Some information about the mental state of these patients during the follow-up period was gathered, but this was self-report information and possibly not so reliable, and we decided not to use it in this study. This was not a follow-up study in its truest sense: the assessments were made only twice – at baseline and 7 years later. Thus, the mental state of the participants during the intervening period is obscure, decreasing slightly the credibility of the study. It is strength of the study that the sample was fairly large, and that it was a follow-up study with a wide range of primary care patients.

Implications

The Depression Scale is not only an easy-to-use screening instrument, it also appears to be a reasonably good predictor for a depressive episode years ahead. It seems to work well with patients who have vague psychiatric symptoms, as is often the case in primary healthcare. Some of its items have a better ability to recognise or to predict depression than others; this suggests the possibility of creating an even shorter version of this scale.