In all, 190 stroke patients were approached for participation. Seventy-five of these were excluded: 21 had recurrent stroke, 17 had cognitive impairment, 17 had aphasia, 10 were < 45 years, and 10 had a stroke duration > 2 years (Fig 1). After applying an exclusion criteria, 115 stroke patients were enrolled, there were 63 males (54.8%) and 52 females (45.2%), with a mean age of 64 + 10 years (min, max: 45,88). The majority had graduated primary school, followed by lower-secondary school and upper-secondary school. The comorbid illnesses found were, in descending order of frequency, hypertension, dyslipidemia, diabetes mellitus, and heart disease. The median duration of stroke was 59 days. Most patients (81.7%) suffered from ischemic stroke, and left-side weakness was dominant (61%). Most patients (65.2%) were recruited from inpatient rehabilitation.
All patients were administered the PHQ- 9 as the index test. The reference standard was a psychiatric interview with diagnoses based on the DSM-5 criteria on the same day. This was administered blind to, and regardless of, scores on the index test. The mean PHQ-9 score was 5.2 + 4.8. According to the DSM-5 criteria, 23 patients (20%) had PSD, while 92 (80%) were normal. In the PSD group, 8 (6.9%) were diagnosed with MDD, 2 (1.7%) had depressive disorder not otherwise specified, and 1 (0.9%) had other specified depressive disorder and 12 patients (10.5%) had adjustment disorder with depressed mood.
The demographic characteristics of the normal and depression groups revealed no statistically significant (age, gender, education level, risk factors, median duration after stroke, pathology of stroke, side of weakness, and settings of the patients). However, the MRS and the median PHQ-9 scores of the groups differed. MRS scores of 0–3 were defined as no disability, while an MRS score > 3 was defined as disability; more stroke patients were disabled in the depression group (78%) than in the normal group (55.4%).
Reliability and item analysis
As shown in Table 1, the highest mean score of the nine PHQ-9 items was found for item 3 (“trouble falling or staying asleep, or sleeping too much”). Item 9 (“thoughts that you would be better off dead or of hurting yourself”) had the lowest score. As to the internal consistency of the PHQ-9, Cronbach’s alpha was 0.78. All items, if deleted, would consistently decrease the total scale alpha. The least item-total correlation was item 5 poor appetite or overeating.
Table 1 Mean score, standard deviation and internal reliability score for the PHQ-9 score
PHQ-9 items
|
mean
|
Standard Deviation
|
Corrected Item-Total Correlation
|
Cronbach's Alpha if Item Deleted
|
1. Little interest or pleasure in doing things
|
0.72
|
0.881
|
0.612
|
0.708
|
2. Feeling down, depressed, or hopeless
|
0.64
|
0.926
|
0.516
|
0.723
|
3. Trouble falling or staying asleep, or sleeping too much
|
1.11
|
1.256
|
0.404
|
0.749
|
4. Feeling tired or having little energy
|
0.68
|
0.984
|
0.321
|
0.755
|
5. Poor appetite or overeating
|
0.47
|
0.955
|
0.199
|
0.773
|
6. Feeling bad about yourself – or that you are a failure
|
0.71
|
1.015
|
0.612
|
0.704
|
7. Trouble concentrating on things
|
0.27
|
0.641
|
0.345
|
0.749
|
8. Moving or speaking so slowly that other people have noticed
|
0.35
|
0.731
|
0.555
|
0.722
|
9. Thoughts that you would be better off dead or of hurting yourself
|
0.25
|
0.662
|
0.525
|
0.729
|
Validity analysis
The performance of the PHQ-9 against the diagnosis of depressive disorder by the DSM-5 criteria for depressive disorder as a criterion standard was examined. According to the DSM-5 criteria, 23 patients (20%) met the diagnosis of post-stroke depression. The median PHQ-9 score for the depression group was 10 (IQR25,75: 7,15) whereas the median score of the normal group was 4 (IQR25,75: 0.5,5.75). These median PHQ-9 score showed statistically different between 2 groups.
Table 2 The performance of different PHQ-9 cut-off scores in detecting depression
Score
|
Sensitivity (%) (95%CI)
|
Specificity (%)(95%CI)
|
Positive predictive value (%)(95%CI)
|
Negative predictive value (%)(95%CI)
|
Positive likelihood ratio (95%CI)
|
Negative likelihood ratio (95%CI)
|
Accuracy (95%CI)
|
Youden’s index
|
The algorithm-based diagnosis
|
>10
|
34.8
(16.4, 57.3)
|
97.8
(92.4, 99.7)
|
80.0
(47.6, 94.6)
|
85.7
(81.6, 89.0)
|
16.0
(3.6, 70.3)
|
85.7
(81.6, 89.0)
|
85.2
(77.4, 91.2)
|
--------------
|
The summed item based diagnosis
|
>5
|
91.3
(71.9, 98.9)
|
65.2
( 54.6, 74.8)
|
39.6
(32.6, 47.2)
|
96.8
(88.8, 99.1)
|
2.62
(1.9, 3.6)
|
0.13
(0.04, 0.5)
|
70.4
(61.2, 78.6)
|
0.565
|
>6
|
87.0
(66.4, 97.2)
|
75.0
(64.9, 83.4)
|
46.5
(37.1, 56.2)
|
95.8
(88.8, 98.5)
|
3.5
(2.4, 5.1)
|
0.2
(0.1, 0.5)
|
77.4
(68.6, 84.7)
|
0.620
|
>7
|
78.3
(56.3, 92.5)
|
81.5
(72.1, 88.8)
|
51.4
(39.6, 63.1)
|
93.8
(87.3, 97.0)
|
4.2
(2.6, 6.8)
|
0.3
0.1, 0.6
|
80.9
(72.5, 87.6)
|
0.598
|
>8
|
65.2
(42.7, 83.6)
|
83.7
(74.5, 90.6)
|
50.0
(36.6, 63.4)
|
90.6
(84.5, 94.4)
|
4.0
(2.3, 6.9)
|
0.42
(0.2, 0.7)
|
80.0
(71.5, 86.9)
|
0.489
|
>9
|
56.5
(34.5, 76.8)
|
90.2
(82.2, 95.4)
|
59.1
(41.4, 74.7)
|
89.3
(83.8, 93.0)
|
5.8
(2.8, 11.8)
|
0.5
(0.3, 0.8)
|
83.5
(75.4, 89.7)
|
0.467
|
>10
|
52.2
(30.59, 73.2)
|
94.6
(87.7, 98.2)
|
70.6
(48.4, 85.9)
|
88.8
(83.7, 92.4)
|
9.6
(3.7, 24.5)
|
0.5
(0.3, 0.8)
|
86.1
(78.4, 91.8)
|
0.467
|
The validity of the index test PHQ-9, using algorithm-based diagnosis, had a sensitivity of 34.8%, specificity of 97.8%, positive predictive value (PPV) of 80%, and negative predictive value (NPV) of 85.7%, and positive likelihood ratio of 16.0 (table 2). When using the summed-scored based diagnosis, a sensitivity, specificity, PPV, NPV, and likelihood ratio of different PHQ-9 thresholds in diagnosing PSD showed in table 2. At the cutoff score of 6 showed the highest Youden’s index. This cutoff score had a sensitivity of 87.0 % (95% CI=66.4, 97.2), specificity of 75.0% (95% CI=64.9, 83.4), positive predictive value of 46.5% (95% CI=37.1, 56.2), negative predictive value of 95.8% (95% CI=88.8, 98.5), positive likelihood ratio of 3.5 (95% CI=2.4, 5.1), and negative likelihood ratio of 0.2 (95%CI=0.1, 0.5). The ROC curve illustrates that the PHQ-9 performed well in identifying patients with PSD (figure 2). The area under the curve (AUC) in our study was 0.87 (95%CI=0.78,0.96) which demonstrated good discrimination.