validation of a screening instrument for postpartum depression in southern brazil

Validation of a screening instrument for postpartum depression inSouthern Brazil

CARLOS ZUBARAN1, KATIA FORESTI2, MARINA VERDI SCHUMACHER2,

ALINE LUZ AMORETTI2, LUCIA CRISTINA MULLER3, MARIANA ROSSI THORELL3,

GILLIAN WHITE4,5, & JOSE MAURO MADI6

1Department of Psychiatry, Blacktown Hospital, West Sydney Area Health Service, School of Medicine, University of Western

Sydney, Australia, 2Department of Biomedical Sciences, University of Caxias do Sul, Brazil, 3Department of Clinical

Medicine, Santa Casa de Misericordia Hospital, Brazil, 4Directorate of Education and Training, Ministry of Health,

Sultanate of Oman, 5School of Health Sciences, Massey University, New Zealand, and 6Department of Clinical and Surgical

Medicine, University of Caxias do Sul, Brazil

(Received 8 November 2008; revised 5 July 2009; accepted 13 July 2009)

AbstractObjectives. The objectives of this investigation was to assess the prevalence of postpartum depression in a sample of 101women and to validate a Portuguese version of the Postpartum Depression Screening Scale (PDSS) in Southern BrazilMethods. Research volunteers completed the PDSS and underwent an assessment based on the Structured ClinicalInterview for DSM-IV disorders (SCID). Parameters under investigation included the demographic characteristics of thesample, internal structure, and discriminant validity.Results. All questions in the Portuguese version of the PDSS attained significant Cronbach’s alpha of 0.62. The factorialanalysis of the Portuguese version of PDSS identified one principal factor that contributed 38.8% of the variance. The bestcut-off score for the Portuguese version of the PDSS was a score of 81, which accounted for a sensitivity of 89% and aspecificity of 72% in this sample.Conclusions. The Portuguese version of the PDSS demonstrated sound psychometric properties. The results of the factorialanalysis also demonstrated that the Portuguese version of the PDSS assesses postpartum depressive disorders in a coherentand integrated manner. The original English version of the PDSS was successfully adapted to Portuguese.

Keywords: Postnatal depression, puerperal disorders, questionnaires, Brazil

Introduction

Mood disturbances represent the most frequent form

of maternal psychiatric morbidity in the postpartum

period [1]. Postpartum depression (PPD) has been

identified as a major public health concern across

different cultures [2], taking into account that it

affects on average 13% of mothers [3]. The diagnostic

criteria of the Diagnostic and Statistical Manual of

Mental Disorders IV (DSM-IV) for depression during

the postpartum period requires an onset that occurs

during the first 4 weeks after birth [4]. By contrast, the

great majority of the postpartum dysphorias begin

within 6 weeks after childbirth [5].

Given that the consequences of PPD affect not only

the mother but also the infant and other family

members [6,7], it is unfortunate that PPD is infre-

quently diagnosed and properly treated [8]. The

symptoms of PPD can remain undetected as they can

resemble normal postpartum adjustment [9]. More-

over, the lack of clear information and appropriate

training programs for the health professionals often

result in an imprecise diagnosis [8].

PPD is a worldwide mental disorder [10], and

transcultural comparisons of depressive symptoma-

tology demonstrate that the symptoms and the

prevalence rates of PPD tend to vary according to

specific cultural backgrounds [11], even when the

same screening instrument is used [12]. It has been

observed that factors such as socio-economic status,

biological vulnerability factors, stigma attached to

mental health disorders, and levels of social support

Correspondence: Carlos Zubaran, University of Western Sydney, School of Medicine, Sydney, Australia. E-mail: [email protected]

Journal of Psychosomatic Obstetrics & Gynecology, December 2009; 30(4): 244–254

ISSN 0167-482X print/ISSN 1743-8942 online � 2009 Informa UK Ltd.

DOI: 10.3109/01674820903254724

J Ps

ycho

som

Obs

tet G

ynae

col D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of

Auc

klan

d on

11/

06/1

4Fo

r pe

rson

al u

se o

nly.

may influence expression of depressive symptoms and

prevalence rates of PPD [2].

Brazil has been considered a multifaceted country

in which several ethnic groups congregate [13]. A

combination of cultural factors such language, tradi-

tions, beliefs, values, and socio-economic status

interact regionally and nationally as the defining

social forces in Brazilian society. The discrepancies

observed between the Northeast – a region affected by

limited food and water resources and deficient

educational and health care systems – and the more

developed Southern region form one of the most

conspicuous national disparities in the Brazil [14].

These regional variances are of such magnitude that it

has been considered ‘‘disputable’’ whether screening

instruments for mental disorders ‘‘can be applied in

different regions and cities of the country without the

interference by local culture’’ [14].

In fact, dissimilar prevalence figures of postnatal

depression have been documented in distinct regions

in Brazil – even when the same screening tool was

used – varying from 12% in Rio de Janeiro [15] and

13.4% in Brasilia [16] to 19.1% in Pelotas [17] and

20.7% in Porto Alegre [18], being the two latter cities

located in Rio Grande do Sul, the Southernmost

Brazilian state. Studies demonstrated high rates of

depressive symptoms among Brazilian women of low

income and non-White skin color during the first 6

months after delivery [15,18,19]. Previous research

evidence had revealed that the prevalence of PPD in

Brazil was highest in Porto Alegre [20].

The routine use of screening scales for the purpose

of identifying characteristic or suggestive symptoms of

depression is an effective, simple, and economical way

to assess women at risk. There are several currently

available self-report measures that have been used to

evaluate depressive symptoms [21], and the Post-

partum Depression Screening Scale (PDSS) [22] has

been widely used with postpartum samples in recent

studies. Moreover, there has been a proportional

increase in the number of studies designed to test and

adapt screening instruments for a series of mental

disorders to different countries and languages [23–

25]. Essentially, a translation process of a psycho-

metric instrument should guarantee linguistic and

cultural equivalence to the original document [26].

Preliminary data of a Portuguese version of the PDSS

as developed via the current research protocol has been

presented previously [27]. Data from a distinct

Portuguese version of PDSS validated in the Northeast

region of Brazil was published subsequently [28]. In the

later study the authors found that ‘‘there were no ideal

translations for a few words’’ because of a lack of

equivalent concepts in Portuguese, whereas in other

adaptations the authors resorted to ‘‘more idiomatic

expressions related to Brazilian culture’’ [28]. Despite

the fact that Brazil has only one official language, there

are significant regional differences of ‘‘local idioms

and accents’’ [14]. Taking into consideration the

significant role that cultural variations may play in the

translation and adaptation of screening and assess-

ment tools, this investigation was conducted to

validate a Portuguese version of the PDSS developed

in Southern Brazil as well as to assess PPD in a

sample of this region.

Material and methods

Pilot study

A pilot study was developed prior to the research

described herein. The conditions in place during the

pilot trial were similar to the ones observed during

the definitive phase of this research. During the pilot

phase, interviewers were uniformly trained to con-

duct interviews with Portuguese version of the

Structured Clinical Interview for DSM-IV Axis I

Disorders (SCID) [29] and with other questionnaires

used in this research protocol. In the Pilot Study, 10

volunteers were assessed. The mean age of this group

was 25.4 years of age. Two participants presented

scores above the cut-off threshold of the Portuguese

version of the EPDS. These women were subse-

quently referred to the Outpatient Psychology Unit of

the UCS for further clinical evaluation. Six inter-

viewers participated in an interrater reliability assess-

ment exercise, which involved the application of the

SCID only. A Kappa of 0.8 was observed among

interviewers. Data from this phase of the study were

not included in this analysis.

Sample recruitment

The study investigated a sample of 101 volunteers,

who underwent deliveries in the General Hospital of

the University of Caxias do Sul (UCS) in Southern

Brazil, which is a higher education training facility

and a regional reference center in Obstetrics.

Research volunteers were recruited via word of

mouth by members of the research team in the days

following delivery while still in hospital. A complete

description of the research protocol was given to

prospective participants. Inclusion criteria were as

follows: (a) patients needed to be available to be

interviewed between the second and the twelfth week

of the postpartum period; (b) patients had to

demonstrate sufficient cognitive capacity to under-

stand and complete the questionnaire as determined

by standardized questions about the purpose of and

the procedures involved in this research; (c) the

mother must have delivered a live and healthy baby,

and (e) volunteers had to complete the informed

consent to participate in this research protocol.

Women who were already being treated for depres-

sion were considered ineligible for the study (N¼ 3).

Volunteers who fulfilled inclusion criteria were

interviewed once at a convenient time in their own

home, during the period encompassed between the

Postnatal depression in Southern Brazil 245

J Ps

ycho

som

Obs

tet G

ynae

col D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of

Auc

klan

d on

11/

06/1

4Fo

r pe

rson

al u

se o

nly.

second and the twelfth week of the postpartum

period. All of the patients who agreed to participate

in this study received assistance from the Brazilian

public health system. All research participants spoke

Portuguese as their native language.

Participants were informed that they could volun-

tarily terminate their participation in the study at any

time without any consequence to themselves or to the

quality of their health care. All information obtained

from research volunteers was treated as confidential.

All patients diagnosed with depressive disorder

according to the SCID interview were promptly

referred to the mental health outpatient clinic of the

Department of Psychology at UCS.

Informed consent

This study was endorsed by UCS Ethics and

Research Committee. All volunteers signed a consent

form to declare a voluntary agreement with all

procedures involved in this project. Minors had to

have their consent endorsed by their parents. Taking

into account that a substantial proportion of the

sample was illiterate or semi-illiterate, all patients

completed the Portuguese version of PDSS under

minimal guidance by trained examiners, who fol-

lowed standardized instructional procedures.

Statistical analysis

Measures of central distribution and dispersion of

data were calculated for all demographic character-

istics. The analysis of internal consistency of the

Portuguese version of the PDSS developed in

Southern Brazil was determined by generating

Cronbach’s alpha indices for each question and area

of this assessment tool. The factorial analysis of the

same scale was conducted by Maximum Likelihood

Factor analysis with Varimax rotation. Application of

the scree test was performed to identify the most

meaningful factorial structure. In addition, a Relative

Operating Characteristic (ROC) analysis, which

considers both sensitivity and specificity factors,

was conducted to establish the best cut-off value.

Optimal cut-off scores and their limitrophe counter-

parts were tested in 26 2 tables for case/noncase

thresholds having the SCID results as a benchmark.

Sensitivity and specificity figures as well as accuracy,

positive and negative predictive values were explored

for each cut-off value. Finally, one-sample w2 tests were

used to assess whether the proportions (cases: non-

cases) associated with the categories were significantly

different from hypothesized proportions. Statistical

analyses were conducted via SPSS1 software.

The postpartum Depression Screening Scale

The PDSS is a self-administered, 35-item Likert scale

that usually requires between 5 and 10 min to be

completed. It was developed to assess the presence and

degree of symptoms of PPD as a screening tool

[10,30]. The PDSS assesses the following seven

dimensions: sleeping/eating disturbances; anxiety/in-

security; emotional lability; mental confusion; loss of

self; guilt/shame; and suicidal thoughts.

The selection of PDSS questions was based on a

series of studies developed by one of the authors of

this psychometric tool [30–32]. Mothers are re-

quested to circle the answer that best describes how

they have felt over the past 2 weeks on a 5-point

Likert scale. Women are asked to indicate their

degree of disagreement or agreement with each item

on a range of strongly disagree (1) to strongly agree

(5). Total score for the PDSS ranges from 35 to 175.

Adaptation and translation procedures

The English version of the PDSS was translated into

Portuguese taking into account semantic, idiomatic,

experiential, cultural, and conceptual equivalence

between the source and the target instruments [33].

Two investigators proficient in both Portuguese and

English developed the final Portuguese version of

the PDSS. Each investigator conducted the trans-

lation and adaptation of the questionnaires from

one language to the other (translation and back-

translation). The final adapted version of the instru-

ment was established by a committee of specialists,

by comparing both the translation and back-transla-

tion of each instrument. This committee consisted of

professionals fully cognizant of the subject under

investigation. Many of them were versed in both

languages. The various drafts of the PDSS, in each

language, were progressively improved by using

relevant information obtained from pilot tests with

samples of patients and professionals. Examples of

questions from each dimension of the Portuguese

version of the PDSS are presented in Appendix 1.

Copies of the Portuguese version of the PDSS can be

obtained from the mailing author upon request.

The translation of the PDSS from English into

Portuguese followed the same principles employed in

the validation process of other questionnaires con-

ducted by the same research group in Brazil [34].

This methodology has analogies with the technique

involved in the conception of the Spanish version of

the PDSS [10].

Results

Sociodemographic characteristics

The mean age of the sample was 25.6 (SD¼ 7.04)

with ages ranging from 14 to 42 years. Seventy-seven

percent of the women were classified by interviewers

as ‘‘White’’ as defined by Brazilian census. As for

marital status, 39% of the mothers had de facto

relationships, 37% were married, 22% were single,

246 C. Zubaran et al.

J Ps

ycho

som

Obs

tet G

ynae

col D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of

Auc

klan

d on

11/

06/1

4Fo

r pe

rson

al u

se o

nly.

and 2% were divorced. In terms of educational

status, two participants had a university degree (2%);

six participants had incomplete university education

(5.9%); 21 participants (20.8%) had completed high

school; 26 mothers (25.7%) had partial high school

education; six mothers had completed primary

school; and 40 research participants (39.6%) had

not completed primary school.

Considering different forms of delivery, 66%

delivered vaginally, whereas 34% had cesarean

delivery. Thirty-nine percent were having their first

child; 32% and 30% were having their second and

third deliveries or subsequent children, respectively.

Fourteen percent of the women presented a previous

history of miscarriage. Fifteen percent of women

were classified as current smokers, whereas 2% of

participants reported regular alcohol consumption.

Twenty-two percent had developed a disease process

during pregnancy, whereas 6% presented complica-

tions during delivery. Seventeen percent of mothers

had a previous history of depression and 57% had a

family history of depression.

Internal structure information

Reliability data. The internal consistency (correlation

of the items to the total score) of the Portuguese

version of the PDSS is shown in Table I. All

questions items included in the PDSS attained

significant Cronbach’s alpha of 0.62 or superior.

The different domains of the Portuguese version of

the PDSS presented Cronbach’s alphas ranging from

0.72 in the area related to anxiety and insecurity area

to 0.86 in the area related to suicidal thoughts. The

PDSS overall Cronbach’s alpha totaled 0.95.

Factorial analysis. Maximum likelihood factor analy-

sis with Varimax rotation was conducted on the

ratings given by the 101 respondents to the 35-item

postnatal depression screening scale. An initial

exploratory factor analysis yielded eight factors with

eigenvalues exceeding unity, explaining 72% of the

variance. As demonstrated in Figure 1, application of

the scree test suggested that a single factor seemed

warranted. The one factor solution appeared the

most tenable option, given that factor one contrib-

uted with 38.8% of the variance. Item loadings for

each of the 35 questions of the Portuguese version of

the PDSS are shown in Table I.

Validity results. A ROC analysis was conducted to

establish the best cut-off score of the Portuguese

version of the PDSS, taking into consideration both

sensitivity and specificity factors. According to this

calculation, the best cut-off value was a score of 81

(AUC¼ 0.88); SE¼ 0.04; 95% CI¼ 0.79–0.96),

which accounts a sensitivity of 89% and a specificity

of 72% in this sample. A ROC curve is obtained by

plotting sensitivity against the false-positive rate for

all possible cut-off points of the screening instru-

ment. The curve represents the performance of the

instrument in discriminating between ‘‘cases’’ and

‘‘noncases’’ across the total spectrum of morbidity,

and the area under the ROC curve (AUC) can be

used as an index of the discriminating ability of a

screening instrument [35]. The limitrophe scores of

73 and 75 were also tested but presented unfavorable

outcomes compared with the proposed cut-off of 74.

As demonstrated in Table II, an increase of the

proposed cut-off to 81 achieved the highest specifi-

city and positive predictive value without a significant

reduction in sensitivity.

Prediction of group membership. In order to test the

discriminative performance of the SCID to differ-

entiate between depressed and nondepressed wo-

men, a one-sample w2 test was conducted. The

results revealed a significant difference between the

two proportions as expected, w2 (1, N¼ 101)¼68.21, p5 0.001. With the purpose of testing the

hypothesis of unequal proportions between de-

pressed and nondepressed women according to

different PDSS cut-off scores as the distinguishing

criteria, a one-sample w2 test was applied. The results

of the analysis having 74 as the PDSS cut-off score

were not significant, w2 (1, N¼ 101)¼ 1.2, p¼ 0.27.

A similar test having the PDSS cut-off score of 81

yielded a significant difference between the two

groups, w2 (1, N¼ 101)¼ 0.94, p¼ 0.05.

Nine women were diagnosed as having PPD

according to SCID criteria. In terms of the screening

performance, the PDSS with a cut-off score of 74

detected 45 positive cases (44.6%) positive cases,

whereas a cut-off of 81 detected 34 positive cases

(33.7%) as demonstrated in Table III. In order to

investigate whether the research participants differ-

entiated by the SCID as depressed or nondepressed

had a matching classification status according to the

PDSS cut-off scores of 74 and 81, a two-way

contingency table analysis was produced. The

screening results according to a cut-off of 74 and

the SCID assessments were found to be significantly

related, Pearson w2 (2, 101)¼ 12.3, p5 0.001

Cramer’s V¼ 0.35. The results generated by a cut-

off of 81 also yielded a significant relationship with

the results obtained by the SCID assessment, with

Pearson w2 (2, 101)¼ 13.49, p5 0.001, Cramer’s

V¼ 0.37.

Discussion and conclusion

The demographic characteristics of this sample were

heterogeneous and represented the usual variations

in clinical presentation at a population level. The age

range varied within exceptional limits of childbearing

years. The sample also presented a diverse nature of


J Ps

ycho

som

Obs

tet G

ynae

col D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of

Auc

klan

d on

11/

06/1

4Fo

r pe

rson

al u

se o

nly.

Table I. The Portuguese version of the Postpartum Depression Screening Scale item loadings (factorial analysis) (n¼101).

Dimension (item)

Response percentages*

Mean

Standard

deviation

Correlation

with

dimension

Dimension

alpha

reliability

if item

deleted

Dimension

alpha

reliability

Factor

loading1 2 3 4 5

Sleeping/eating disturbances 0.86

Had trouble sleeping

even when my baby

was asleep (1)

44.6 22.8 3 9.9 19.8 2.38 1.59 0.66 0.84 0.66

Lost my appetite (8) 50.5 26.7 2 14.9 5.9 1.99 1.29 0.57 0.86 0.72

Woke up on my own

in the middle of the night

and had trouble getting

back to sleep (15)

40.6 23.8 4 22.8 8.9 2.36 1.43 0.75 0.81 0.76

Tossed and turned for

a long time at night

trying to fall

asleep (22)

42.6 25.7 4 17.8 9.9 2.27 1.42 0.79 0.8 0.82

Knew I should eat but

I could not (29)

54.5 19.8 2 13.9 9.9 2.05 1.42 0.64 0.84 0.66

Anxiety/insecurity 0.72

Got anxious over even

the littlest things that

concerned my baby (2)

13.9 19.8 7.9 39.6 18.8 3.30 1.35 0.3 0.75 0.63

Felt really overwhelmed (9) 47.5 26.7 3 18.8 4 2.05 1.28 0.59 0.64 0.70

Felt like I was jumping out

of my skin (16)

38.6 22.8 7.9 21.8 8.9 2.40 1.41 0.61 0.62 0.69

Felt all alone (23) 37.6 19.8 5.9 25.7 10.9 2.52 1.48 0.56 0.65 0.63

Felt like I had to keep

moving or pacing (30)

52.5 27.7 3 10.9 5.9 1.90 1.24 0.39 0.71 0.67

Emotional liability 0.8

Felt like my emotions were

on a roller coaster (3)

24.8 20.8 6.9 31.7 15.8 2.93 1.47 0.61 0.75 0.7

Was scared that I would

never be happy again (10)

55.4 26.7 3 11.9 3 1.80 1.14 0.47 0.79 0.72

Cried a lot for no real

reason (17)

45.5 21.8 3 22.8 6.9 2.24 1.41 0.53 0.78 0.65

Have been very irritable (24) 26.7 19.8 6.9 28.7 17.8 2.91 1.51 0.67 0.73 0.71

Felt full of anger, ready

to explode (31)

47.5 20.8 5 18.8 7.9 2.19 1.40 0.64 0.74 0.72

Mental confusion 0.82

Felt like I was losing my

mind (4)

47.5 21.8 8.9 17.8 4 2.09 1.28 0.57 0.79 0.71

Could not concentrate on

anything (11)

39.6 38.6 3 11.9 6.9 2.08 1.24 0.68 0.76 0.79

Thought I was going crazy (18) 59.4 26.7 1 10.9 2 1.69 1.06 0.58 0.79 0.80

Had a difficult time making

even a simple decision (25)

41.6 20.8 3 25.7 8.9 2.40 1.46 0.57 0.79 0.62

Had difficulty focusing on

a task (32)

43.6 26.7 5.9 14.9 8.9 2.19 1.37 0.66 0.76 0.72

Loss of self 0.78

Was afraid that I would

never be my normal

self again (5)

41.6 26.7 1 18.8 11.9 2.33 1.47 0.37 0.84 0.69

Felt as though I had become

a stranger to myself (12)

56.4 32.7 2 7.9 1 1.64 0.93 0.66 0.71 0.66

Did not know who I was

anymore (19)

62.4 27.7 3 5.9 1 1.55 0.89 0.65 0.72 0.79

Felt like I was not normal (26) 56.4 31.7 1 6.9 4 1.70 1.06 0.7 0.69 0.80

Did not feel real (33) 57.4 26.7 3 9.9 3 1.74 1.10 0.55 0.74 0.79

Guilt/shame 0.78

Felt like I was not the mother

I wanted to be (6)

48.5 21.8 10.9 17.8 1 2.01 1.19 0.55 0.75 0.67

(continued)


J Ps

ycho

som

Obs

tet G

ynae

col D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of

Auc

klan

d on

11/

06/1

4Fo

r pe

rson

al u

se o

nly.

delivery types and progeny. The sample was, there-

fore, constituted of typical women seeking obstetric

attention in a public health service in Southern

Brazil. The Portuguese adaptation of the PDSS was

well accepted, considering that no participants

refused to complete the interview after they had

signed the consent form. The data-gathering process

was conducted by a limited number of interviewers

and data transfer to a desktop environment was also

very efficient.

The Portuguese version of the PDSS proved to

have sound psychometric properties. Internal con-

sistency data revealed significant Cronbach’s coeffi-

cients in all seven areas of the PDSS. The internal

consistency of each dimension assessed by the

Portuguese version of the PDSS reflects the accuracy

of the measurement process. The results of the PDSS

factorial analysis demonstrated adequate homogene-

ity and coherence of all items included in this

subscale, which represents an integrated assessment

of postpartum depressive disorders.

Although eight factors were extracted in an

exploratory factor analysis, the scree test revealed

one main factor. Different criteria have been pro-

posed for establishing the number of factors to

extract based on the magnitudes of the eigenvalues.

On criterion is to retain all factors that have

eigenvalues greater than the unity, whereas another

method is to examine the scree plot and to retain

factors with eigenvalues in the sharp descent part of

the plot before the eigenvalues start to plateau [36].

In the initial study of the PDDS, which revealed a

seven-factor structure, a confirmatory factor analysis

Table I. (Continued).

Dimension (item)

Response percentages*

Mean

Standard

deviation

Correlation

with

dimension

Dimension

alpha

reliability

if item

deleted

Dimension

alpha

reliability

Factor

loading1 2 3 4 5

Felt like so many mothers

were better than me (13)

59.4 19.8 4 11.9 5 1.83 1.24 0.66 0.71 0.74

Felt guilty because I could

not feel as much love

for my baby as I should (20)

66.3 22.8 1 7.9 2 1.56 0.99 0.54 0.75 0.71

Felt like I had to hide what

I was thinking or

feeling towards the

baby (27)

59.4 23.8 1 10.9 5 1.78 1.20 0.38 0.80 0.69

Felt like a failure as a

mother (34)

67.3 22.8 2 7.9 0 1.50 0.88 0.75 0.7 0.78

Suicidal thoughts 0.86

Have thought that death

seemed like the only

way out of this living

nightmare (7)

78.2 12.9 2 5.9 1 1.39 0.87 0.73 0.82 0.82

Started thinking that

I would be better

off dead (14)

76.2 16.8 1 4 2 1.39 0.86 0.89 0.78 0.90

Wanted to hurt myself (21) 73.3 22.8 1 2 1 1.35 0.7 0.53 0.87 0.63

Felt that my baby

would be better off

without me (28)

77.2 19.8 0 1 2 1.31 0.71 0.46 0.88 0.81

Just wanted to leave

this world (35)

72.3 19.8 0 5.9 2 1.46 0.92 0.92 0.8 0.85

Number in parenthesis refers to the item number in the PDSS-Portuguese Version.

*Rating scale: 1¼ strongly disagree; 2¼disagree; 3¼neither disagree nor agree; 4¼ agree; 5¼ strongly agree.

Figure 1. Scree plot of the eigenvalues.


J Ps

ycho

som

Obs

tet G

ynae

col D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of

Auc

klan

d on

11/

06/1

4Fo

r pe

rson

al u

se o

nly.

approach was used instead. The validation study of

the Turkish version of the PDSS revealed only six

factors with Eigenvalues superior to one [37]. The

fact that the eingenvalues ranged from 1.15 to 11.46

strongly suggest that the scree test, not reported in

the Turkish study, would have probably indicated a

different number of factors to account for the

variance. The previous study of the Portuguese

version of the PDSS did not demonstrate a factorial

extraction of factors [28]. A factorial analysis based

on the results of the scree test yields accurate results

more often than the extraction method based on

eigenvalues with values grated than the unity [36].

The PDSS score of 81 and the SCID assessment

were both able to effectively differentiate between

depressed and non-depressed women. This differ-

entiation could not be achieved with the lower PDSS

cut-off score of 74. A significant correlation was

observed between the cut-off score of 81 and the

results produced by the SCID assessment.

Versions of the PDSS have been adapted to Thai

[38], Turkish [37], Spanish [10], and Portuguese

[28]. The original English version of the PDSS has

also been tested among Native American English

[39] and First Nations and Metis women [40]. The

cut-off score of 80 was recommended for the

screening of major PPD with the Original English

version of the PDSS [41]. The cut-off score for major

PPD was not established in the validation process of

the Spanish version of the PDSS, in which different

cut-off scores were conflated to screen for significant

symptoms of PPD as well [10]. The proposed

screening cut-off score for major PPD in the Thai

version was 90 [38] and the study with First Nations

and Metis women recommended a respective cut-off

score of 71.5 [40]. The Portuguese version of the

PDSS previously validated in Brazil proposed a

screening score of 102 for postpartum major depres-

sion, which yielded a sensitivity of 93.8% and a

specificity of 93.3% [28]. This cut-off score contrasts

notably with the score of 81 proposed by this study.

Apart from the regional linguistic and cultural

difference from Northeast to Southern Brazil, there

are additional possibilities to account for this differ-

ence. The previous study conducted in Brazil

evaluated two distinct groups of same size, from the

public and the private health care sector [28], whereas

in this study the sample was recruited from a public

hospital only. The inequalities between the private and

the public health care sector epitomize the abysmal

socio-economic differences in Brazil, where profes-

sionals, and members of their families, are predomi-

nant in private hospitals, whereas workers of lower

income are predominant in public hospitals [42,43]. It

has been reported that the risk of death for patients

treated in the public sector is significantly higher than

in private hospitals in Brazil [44].

In the preceding study in Brazil, 38.3% had

obtained a higher education degree or were still

enrolled in college [28], whereas in our study only

2% of women had a university degree. Therefore,

the sample investigated in the prior study was

significantly more educated and probably more

economically advantaged that the sample investi-

gated in this study. Data from 2000 census con-

ducted by the Brazilian Institute of Geography and

Statistics (IBGE) revealed that only 6.7% of Brazilian

women 25 years of age and older have completed a

Table II. Screening performance of different cut-off scores in the Portuguese Version of PDSS.

PDSS cut-off scores Sensitivity (%) Specificity (%) Positive predictive value (%) Negative predictive value (%) Accuracy (%)

.

.

.74 100 61 20 100 64

.

.

.81 89 72 23 98 73

82 78 74 23 97 74

.

.

.

Table III. Predicted group membership by the results of the SCID

versus the results of the PDSS according to the proposed cut-off

score of 81.

SCID

Positive Negative

PDSS

Positive 8 34 ! Positive

predictive

value¼0.23

Negative 1 58 ! Negative

predictive

value¼0.98

# #Sensitivity¼ 0.89 Specificity¼0.72


J Ps

ycho

som

Obs

tet G

ynae

col D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of

Auc

klan

d on

11/

06/1

4Fo

r pe

rson

al u

se o

nly.

higher education degree [45]. In addition, the

sensitivity the PDSS with the proposed screening

score of 102 dropped considerably from 100% in the

private setting to 80% among patients assisted in the

public system [28]. The positive predictive value

displayed an even more significant reduction from

91.7% to 50%. The authors of the previous study

acknowledged that cut-off scores derived from a

given sample ‘‘may not perform well in other samples

and need to be cross-validated in order to be

accepted for clinical use’’ [28].

Another difference between this study and the

previous study conducted in Brazil is the timeframe

chosen for the screening interview with the PDSS to

be performed. Although in this study the assessment

was conducted from the second to the twelfth week,

in the previous study the assessment was conducted

from the second to the 26th week. It has been

advocated ‘‘the comparison between the results of

different validation studies for a same test requires

caution’’, because several factors may influence the

screening performance of a test such as the pre-

valence of the disease in the sample, the design of the

study, and the study population’s socio-cultural

characteristics [46].

Currently, distinct self-report measures are used to

evaluate depressive symptoms in Brazil. The EPDS,

the most widely used screening scale worldwide for

PPD, has been translated into several languages and

validated in different countries [47]. Previous studies

have reported the performance of EPDS in different

regions in Brazil [15,16,46]. The EPDS has potential

advantages over the PDSS, such as the shorter period

required to apply the former instrument, and the

extensive utilization by investigators and clinicians in

different socioeconomic setting and with a variety of

ethnic groups. On the other hand, the PDSS can be

used to evaluate specific areas of interest, which

facilitates identification of preferential treatment

options [47]. Additional features that may occur in

a mood disorder are not measured by the EPDS,

including loss of control, loneliness, irritability,

obsessive thinking, and difficulty concentrating

[22]. According to the pioneering studies of the

PDSS, the ability of this instrument to detect PPD is

comparable with the Structured Clinical Interview

for DSM-IV Axis I Disorders (SCID) for detecting

major depressive disorders. Additional evidence also

indicates that the PDSS may be more accurate than

the EPDS and Beck Depression Inventory (BDI) at

detecting major and minor depression at 12 weeks

postpartum [22,41,47].

The results generated by this research may reflect

some limitations. The optimal proposed cut-off of 81

produced a low positive predictive value of 23.

However, as reviewed elsewhere, because of the

generally low prevalence of diseases, screening tests

tend to present low positive predictive values, even

when specificity is high [46,48]. In fact, the ROC

analysis identified 74 as a score that would produce a

sensitivity of 100% and a specificity of 61% in this

sample. As advocated elsewhere, a screening test that

fails to identify more than one-fourth of mothers with

PPD is unsatisfactory, because a good screening test

should present high sensitivity in order avoid missing

disease cases, and high specificity, in order to

diminish false-positives that tend to require further

evaluation. There are two ways of increasing the

positive predictive value of a test: (1) increasing the

prevalence of the disease to be screened by a given

instrument and (2) adjusting the cut-off of the

instrument in order to achieve higher specificity

[46]. Further increments of the cut-off would

produce unacceptably low sensitivity.

In addition, considering that PPD is typically

transitory in nature, the period of 4 weeks encom-

passed within the temporal limits of the data

collection in this study (from the 6th to the 10th

postpartum week) may generate inexact prevalence

figures of PPD. The PDSS assesses phenomena

related only to the preceding 2 weeks. In a significant

number of women within an ample age range,

postpartum depressive symptoms persist for months

or even years after delivery [49]. Longitudinal data

demonstrated that almost half of adult and adoles-

cent mothers diagnosed with depression during the

last trimester of pregnancy remained depressed

during the second and third months of the post-

partum period [50]. Women from a low socio-

economic background may be particularly vulnerable

for developing depressive disorders of longer dura-

tion or later onset [51].

Considering that the PDSS was developed to

screen for depressive disorders among adult mothers,

it can be argued that this instrument may perform

differently among minors. Research evidence indi-

cates that adolescent mothers are more likely then

older mothers to be depressed in the postpartum

period [52]. This finding has been replicated by

studies using screening tools usually reserved for

adults, including the EPDS [50] and the BDI [53],

or by specific instruments specifically devised for

youngsters such as the Center for Epidemiological

Studies Depression Scale for Children [54]. Other

studies conducted with different versions of the

PDSS included minors in their sample, including

the Thai version (age range from 15 to 43) [38], the

Spanish version (16–44) [10], the previous Portu-

guese version (15–45) [28], and the study conducted

with native Americans participants (age range from

16 to 40) [39].

In summary, this study demonstrates that the

original English version of the PDSS was successfully

adapted to Portuguese as methodologically demon-

strated. The internal consistency of the Portuguese

version of the PDSS, as indicated by the results of the


J Ps

ycho

som

Obs

tet G

ynae

col D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of

Auc

klan

d on

11/

06/1

4Fo

r pe

rson

al u

se o

nly.

Cronbach’s coefficients, was satisfactory, which

demonstrates that the instrument coherently investi-

gates postpartum depressive disorders as measured

uniformly by its items. This assessment tool was well

accepted by all volunteers. The screening perfor-

mance of this Portuguese version of the PDSS should

be interpreted in the light of the community sample

investigated, which represents a population sample

without any previous documentation of depressive

symptomatology.

Declaration of interest: The authors report no

conflicts of interest. The authors alone are respon-

sible for the content and writing of the paper.

References

1. Dennis C. Can we identify mothers at risk for postpartum

depression in the immediate postpartum period using the

Edinburgh Postnatal Depression Scale? J Affect Disord 2004;

78:163–169.

2. Halbreich U, Karkun S. Cross-cultural and social diversity of

prevalence of postpartum depression and depressive symp-

toms. J Affect Disord 2006;91:97–111.

3. O’Hara M, Swain A. Rates and risks of postpartum depres-

sion: a meta-analysis. Nurs Res 1996;8:37–54.

4. American Psychiatric Association. Diagnostic and statistical

manual of mental disorders. Washington, DC: American

Psychiatric Association; 2000.

5. Sadock BJ, Sadock VA. Kaplan and Sadock’s synopsis of

psychiatry: behavioral sciences/clinical psychiatry. Philadel-

phia, PA: Lippincott Williams & Wilkins; 2007.

6. Chaudron LH, Szilagyi PG, Kitzman HJ, Wadkins H, Conwell

Y. Detection of postpartum depressive symptoms by screening

at well-child visits. Pediatrics 2004;113:551–558.

7. Miller LJ. Pospartum depression. JAMA 2002;287:762–766.

8. Lee DT, Yip AS, Chiu HF, Chung TK. Screening for

postnatal depression using the double-test strategy. Psycho-

som Med 2000;62:258–263.

9. Epperson CN. Postpartum major depression: detection and

treatment. Am Fam Physician 1999;59:2247–2254.

10. Beck CT, Gable RK. Postpartum Depression Screening

Scale – Spanish version. Nurs Res 2003;52:296–306.

11. Jablensky A, Sartorius N, Gulbinat W, Ernberg G. Character-

istics of depressive patients contacting psychiatric services in

four cultures. A report from the who collaborative study on the

assessment of depressive disorders. Acta Psychiatr Scand

1981;63:367–383.

12. Bashiri N, Spielvogel AM. Postpartum depression: a cross

cultural perspective. Psychiatry Update 1999;6:82–87.

13. Zubaran C. The quest for recognition: Brazilian immigrants

in the United States. Transcult Psychiatry 2008;45:590–

610.

14. Carvalho AL, Silva LF, Grisi SJ, Escobar AM. Clinical

indicators of child development in the capitals of nine

Brazilian states: the influence of regional cultural factors.

Clinics 2008;63:51–58.

15. Da Silva VA, Santos AR, Carvalho MS, Martins ML, Teixeira

NA. Prenatal and postnatal depression among low-income

Brazilian women. Brz J Med Biol Res 1998;31:799–804.

16. Santos MFS, Martins FC, Pasquali L. Escalas de auto-

avaliacaode depressao pos-parto: estudo Brasil. Rev Psiquia-

tria Clın (Sao Paulo) 1999;26:32–40.

17. Moraes IG, Pinheiro RT, Silva RA, Horta BL, Sousa PL,

Faria AD. [Prevalence of postpartum depression and asso-

ciated factors]. Rev Saude Publica 2006;40:65–70.

18. Tannous L, Gigante LP, Fuchs SC, Busnello ED. Postnatal

depression in Southern Brazil: prevalence and its demographic

and socioeconomic determinants. BMC Psychiatry 2008;8:1.

19. Faisal-Cury A, Tedesco JJ, Kahhale S, Menezes PR, Zugaib M.

Postpartum depression: in relation to life events and patterns of

coping. Arch Women Ment Health 2004;7:123–131.

20. Almeida-Filho N, Mari Jde J, Coutinho E, Franca JF,

Fernandes J, Andreoli SB, Busnello ED. Brazilian multicentric

study of psychiatric morbidity. Methodological features and

prevalence estimates. Br J Psychiatry 1997;171:524–529.

21. Boyd RC, Le HN, Somberg R. Review of screening

instruments for postpartum depression. Arch Womens Ment

Health 2005;8:141–153.

22. Beck CT, Gable RK. Postpartum Depression Screening Scale:

development and psychometric testing. Nurs Res 2000;49:

272–282.

23. Beaton DE, Bombardier C, Guillemin F, Ferraz MB. Guide-

lines for the process of cross-cultural adaptation of self-report

measures. Spine 2000;25:3186–3191.

24. Bullinger M, Alonso J, Apolone G, Leplege A, Sullivan M,

Wood-Dauphinee S, Gandek B, Wagner A, Aaronson N,

Bech P et al. Translating health status questionnaires and

evaluating their quality: the IQOLA Project approach. Inter-

national Quality of Life Assessment. J Clin Epidemiol 1998;

51:913–923.

25. Wiesinger GF, Nuhr M, Quittan M, Ebenbichler G, Wolfl G,

Fialka-Moser V. Cross-cultural adaptation of the Roland-

Morris questionnaire for German-speaking patients with low

back pain. Spine 1999;24:1099–1103.

26. Beck CT, Bernal H, Froman RD. Methods to document

semantic equivalence of a translated scale. Res Nurs Health

2003;26:64–73.

27. Madi JM, Schumacher M, Muller L, Foresti K, Amoretti A,

Zubaran C. Prevalence of depressive symptoms in the

postpartum period. Poster presented at 51st Brazilian Con-

gress of Gynecology and Obstetrics. Rio de Janeiro, RJ; 2005.

28. Cantilino A, Carvalho JA, Maia A, Albuquerque C, Cantilino

G, Sougey EB. Translation, validation and cultural aspects of

postpartum depression screening scale in Brazilian Portu-

guese. Transcult Psychiatry 2007;44:672–684.

29. Steinberg M. Interviewers guide to the structured clinical

interview for DSM-IV dissociative disorders (SCID-D).

Washington, DC: American Psychiatric Press; 1994.

30. Beck CT. Teetering on the edge: a substantive theory of

postpartum depression. Nurs Res 1993;42:42–48.

31. Beck CT. The lived experiences of postpartum depression: a

phenomenological study. Nurs Res 1992;41:166–170.

32. Beck CT. Postpartum depressed mothers’ experiences inter-

acting with their children. Nurs Res 1996;45:98–104.

33. Koller M, Aaronson NK, Blazeby J, Bottomley A, Dewolf L,

Fayers P, Johnson C, Ramage J, Scott N, West K.

Translation procedures for standardised quality of life ques-

tionnaires: the European Organisation for Research and

Treatment of Cancer (EORTC) approach. Eur J Can

2007;43:1810–1820.

34. Zubaran C, Persch K, Tarso D, Ioppi AE, Mezzich J. The

Portuguese version of the Personal Health Scale: a validation

study in southern Brazil. Clinics 2007;62(4):419–426.

35. Mari JJ, Williams P. A comparison of the validity of two

psychiatric screening questionnaires (GHQ-12 and SRQ-20)

in Brazil, using Relative Operating Characteristic (ROC)

analysis. Psychol Med 1985;15:651–659.

36. Green S, Salkind N. Scales procedures. In: Using SPSS for

Windows and Macintosh: analyzing and understanding data,

4th ed. Upper Sadle River, NJ: Pearson Prentice Hall; 2005.

p 317.

37. Karacam Z, Kitis Y. [The Postpartum Depression Screening

Scale: its reliability and validity for the Turkish population].

Turk Psikiyatri Derg 2008;19:187–196.


J Ps

ycho

som

Obs

tet G

ynae

col D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of

Auc

klan

d on

11/

06/1

4Fo

r pe

rson

al u

se o

nly.

38. Vittayanont A, Liabsuetrakul T, Pitanupong J. Development

of Postpartum Depression Screening Scale (PDSS): a Thai

version for screening postpartum depression. J Med Assoc

Thai 2006;89:1–7.

39. Baker L, Cross S, Greaver L, Wei G, Lewis R. Prevalence of

postpartum depression in a native American population.

Matern Child Health J 2005;9:21–25.

40. Clarke PJ. Validation of two postpartum depression screening

scales with a sample of First Nations and Metis women. Can J

Nurs Res 2008;40:113–125.

41. Beck CT, Gable RK. Further validation of the Postpartum

Depression Screening Scale. Nurs Res 2001;50:155–164.

42. Rocha JS, Simoes BJ. [Study of public and private hospital

care on a population basis, 1986–1996]. Rev Saude Publica

1999;33:44–54.

43. Rocha JS, Simoes BJ, Guedes GL. [Hospital assistance as an

indicator of social inequality]. Rev Saude Publica 1997;31:

479–487.

44. Lucif N Jr, Rocha JS. [Study of inequalities in hospital

mortality using the Charlson comorbidity index]. Rev Saude

Publica 2004;38:780–786.

45. Brazilian Institute of Geography and Statistics (IBGE). [Demo-

graphic census – educational data]. Rio de Janeiro, RJ: Brazilian

Institute of Geography and Statistics (IBGE); 2000.

46. Santos IS, Matijasevich A, Tavares BF, Barros AJ, Botelho IP,

Lapolli C, Magalhaes PV, Barbosa AP, Barros FC. Validation

of the Edinburgh Postnatal Depression Scale (EPDS) in a

sample of mothers from the 2004 Pelotas Birth Cohort Study.

Cad Saude Publica 2007;23:2577–2588.

47. Hanusa BH, Scholle SH, Haskett RF, Spadaro K, Wisner KL.

Screening for depression in the postpartum period: a

comparison of three instruments. J Womens Health (Larchmt)

2008;17:585–596.

48. Fletcher RH, Fletcher SW, Wagner EH. Epidemiologia

clınica: elementos essenciais. Porto Alegre, BR: Editora Artes

Medicas; 1996.

49. Goodman JH. Postpartum depression beyond the early

postpartum period. J Obstet Gynecol Neonatal Nurs 2004;

33:410–420.

50. Figueiredo B, Pacheco A, Costa R. Depression during

pregnancy and the postpartum period in adolescent and adult

Portuguese mothers. Arch Womens Ment Health 2007;10:

103–109.

51. Seguin L, Potvin L, St-Denis M, Loiselle J. Depressive

symptoms in the late postpartum among low socio-economic

status women. Birth 1999;26:157–163.

52. Deal LW, Holt VL. Young maternal age and depressive

symptoms: results from the 1988 national maternal

and infant health survey. Am J Public Health 1998;88:266–

270.

53. Schmidt RM, Wiemann CM, Rickert VI, Smith EO.

Moderate to severe depressive symptoms among adolescent

mothers followed four years postpartum. J Adolesc Health

2006;38:712–718.

54. Hudson DB, Elek SM, Campbell-Grossman C. Depression,

self-esteem, loneliness, and social support among adolescent

mothers participating in the new parents project. Adolescence

2000;35:445–453.

Appendix 1. Examples of questions from each dimension of the Portuguese version of PDSS.

English version (over the last two weeks, I. . .) Portuguese version (durante as duas ultimas semanas, eu. . .)

Dimension (item)

Sleeping/eating disturbances

Had trouble sleeping even when my baby was asleep (1) tive problema com sono mesmo quando o meu bebe estava

dormindo

Lost my appetite (8) perdi meu apetite

Anxiety/insecurity

Felt all alone (23) me sentia completamente sozinha

Felt like I had to keep moving or pacing (30) sentia como se eu tivesse que me manter em movimento ou

andando

Emotional liability

Was scared that I would never be happy again (10) estava assustada que eu nunca seria feliz novamente

Have been very irritable (24) estive muito irritada

Mental confusion

Thought I was going crazy (18) pensava que estava ficando louca

Had a difficult time making even a simple decision (25) tive dificuldade mesmo para tomar uma decisao simples

Loss of self

Felt like I was not normal (26) me sentia como se nao fosse normal

Did not feel real (33) nao me sentia real

Guilt/shame

Felt like so many mothers were better than me (13) sentia como se tantas outras maes eram melhores do que eu

Felt guilty because I could not feel as much love for my baby

as I should (20)

me sentia culpada porque eu nao conseguia sentir tanto amor

pelo meu filho quanto eu deveria

Suicidal thoughts

Wanted to hurt myself (21) quis me machucar

Felt that my baby would be better off without me (28) sentia que meu bebe estaria melhor sem mim


J Ps

ycho

som

Obs

tet G

ynae

col D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of

Auc

klan

d on

11/

06/1

4Fo

r pe

rson

al u

se o

nly.

Current knowledge on this subject

. PPD is a worldwide mental disorder that affects on average 13% of mothers. The symptomatic

manifestation of PPD is significantly influenced by cultural aspects. In Brazil there are significant cross-

regional variations of cultural and socio-economic factors. Different prevalence figures of postnatal

depression have been reported in distinct regions in Brazil. The use of screening tools for the purpose

of identifying symptoms of depression is an effective method for assessing women at risk for PPD. The

PDSS is a 35-item instrument that has been validated in several languages including Portuguese.

What this study adds

. A new a Portuguese version of the PDSS was developed and tested among 101 postpartum women in

Southern Brazil. The results demonstrated that this version possesses sound psychometric properties.

The screening performance of this version demonstrated a significant correlation with the results of

diagnostic interviews conducted according to the Structured Clinical Interview for DSM-IV Axis I

Disorders (SCID).


J Ps

ycho

som

Obs

tet G

ynae

col D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of

Auc

klan

d on

11/

06/1

4Fo

r pe

rson

al u

se o

nly.

validation of a screening instrument for postpartum depression in southern brazil

Documents