Barriers to access to treatment for mothers with postpartum depression in primary health care centers: a predictive model1

Objective to develop a predictive model to evaluate the factors that modify the access to treatment for Postpartum Depression (PPD). Methods prospective study with mothers who participated in the monitoring of child health in primary care centers. For the initial assessment and during 3 months, it was considered: sociodemographic data, gyneco-obstetric data, data on the services provided, depressive symptoms according to the Edinburgh Postpartum Depression Scale (EPDS) and quality of life according to the Short Form-36 Health Status Questionnaire (SF-36). The diagnosis of depression was made based on MINI. Mothers diagnosed with PPD in the initial evaluation, were followed-up. Results a statistical model was constructed to determine the factors that prevented access to treatment, which consisted of: item 2 of EPDS (OR 0.43, 95%CI: 0.20-0.93) and item 5 (OR 0.48, 95%CI: 0.21-1.09), and previous history of depression treatment (OR 0.26, 95%CI: 0.61-1.06). Area under the ROC curve for the model=0.79; p-value for the Hosmer-Lemershow=0.73. Conclusion it was elaborated a simple, well standardized and accurate profile, which advises that nurses should pay attention to those mothers diagnosed with PPD, presenting low/no anhedonia (item 2 of EPDS), scarce/no panic/fear (item 5 of EPDS), and no history of depression, as it is likely that these women do not initiate treatment.


Introduction
Postpartum depression (PPD) is a public health problem worldwide (1) . It is the most common psychiatric condition postpartum (2) and there is extensive material on the degree of disability that it is likely to cause to the mother (3) , its association with the delay in child development and behavior disorders in adult life of the descendants (4) .
In Chile, studies using standardized diagnostic criteria reported a prevalence of PPD of about 20% in the primary health care (PHC) of public health system (5) . In contrast, a study using the Edinburgh Postpartum Depression Scale (EPDS), validated in Chile (6) , indicated that 41.3% of mothers who are assisted in clinics are affected by severe depressive symptoms between 2 and 3 months postpartum (7) , that is, at risk of PPD.
Although a significant proportion of mothers who use the APS are at high risk and the importance of maternal and child health leads to a greater number of visits to health centers in this period, depressive disorders are not usually detected and treated (8) , despite the availability of effective treatments (9) .
Based on that, the Ministry of Health (10) promoted a early detection of PPD, recommending the adoption of the universal screening in the PHC, so that the EPDS is applied by nursing professionals in the follow-up of children and women at postpartum period. However, treatment rates remain low.
In this regard, the national literature has evidenced the presence of barriers to access to health services for depressed mothers and the need for trainnig of human resources in the PHC in order to ensure a greater commitment to the ministerial guidelines and tighter monitoring of women at risk (11) .
It is considered that the construction of a predictive model to identify the factors that modify the access to treatment may be useful in reducing the failures in the treatment of PPD, by focusing on the use of human resources available in the public health system, and specifically, strengthening the role of nurses in detecting PPD during routine examinations.
There are no studies in the local context that have investigated that aspect at present.
The aim of this study was to develop a predictive model to evaluate the factors that modify the access to treatment for PPD in PHC.

Method
This is a prospective cohort study. The sampling consisted of all health units of PHC located in the Metropolitan Region (MR), Chile (n=120 A week later, a structured interview was carried out by phone (initial diagnosis), which assessed: sociodemographic antecedents, gynecological-obstetric and perinatal data, depressive symptoms, according to the Edinburgh Postpartum Depression Scale (EPDS) (7) , confirmation of current diagnosis of Major Depressive Postpartum Episode (PPD), according to the structured psychiatric interview MINI (12) and quality of life, according to the SF-36 Health Status Questionnaire (13) .
The final sample used for collection and analysis of data in this study included only women in which PPD has been confirmed, according to MINI, in the initial diagnosis.

Definition of dependent variable
After three months, the medical records of users with PPD (follow-up evaluation) were reviewed, considering as no access to treatment: if no provision of mental health consultation was recorded in the health unit after the initial diagnosis (dichotomized variable).

Definition of independent variables
To determine the predictors of no access to treatment in women with PPD in PHC, a review of the available literature was performed (14)(15)(16)(17)(18)(19)(20)(21) . Accordingly, the following variables were selected as potential predictors: All variables that were significant with p<0.1 in the univariate analysis, were included in the multivariate model using a backward selection technique (backward), to obtain the most parsimonious multivariate predictive model. The Hosmer-Lemeshow test was used to measure the effectiveness of the predictive model, that is, the matching between the predicted and observed probabilities. To evaluate the discrimination ability of the model, that is, the probability to identify a case of PPD from a couple of observations taken at random, it was used the area under the ROC curve (Receiver Operating Characteristics). Statistical analyzes were performed with Stata 12.0 (22) . All estimates were presented collectively with confidence intervals at 95% (95%CI).

Results
The initial sample consisted of 305 women. In the initial diagnosis, PPD was confirmed in 63 of them (20.7%), which formed the final sample for the analysis.
In the follow-up evaluation, it was possible to access the medical records of all women in the final sample, therefore, there was no loss of data.
As shown in Table 1    In the final model (  Although the latter two predictors were not statistically significant, they were "forced" into the model due to their contribution to a more parsimonious development of the predictive model and also based on the literature, which supported their inclusion (15)(16)(17)21) .

Discussion
This is the first study in the national literature to develop a predictive model to evaluate the factors influencing the access to treatment for PPD in mothers who use PHC. Access to treatment of women with PPD is still very low, despite the existence of universal access and the availability of effective treatments.
According to this study, women who develop PPD and with no access to treatment are those presenting low levels of anhedonia and symptoms of anxiety (panic and fear), and who did not have prior history of treatment due to episodes of depression.  (8) , therefore, it is relevant to public health and to the role played by nursing professionals at postpartum.
However, the practical significance of these findings must be viewed with caution. This study is a secondary analysis of databases on a research that was developed for other purposes, which imposes important limitations: it is likely that eventually significant predictors have not been included, since access to treatment of postpartum depression has been described as a complex phenomenon that involves not easily quantifiable variables such as domestic workload, the ideals of motherhood and the stigma associated with mental health problems (12,17) ; In addition, the analyzes were performed based on a small sample (n=63), which could affect the power of the study.
However, the non-inclusion of variables consideres as difficult to measure ("complex") is related with the development of a pragmatic risk profile, relatively easy to use and which does not require an additional effort from the nursing profesional in PHC. This is not a matter of dismissing important topics for addressing the PPD (and maternal health, in general) such as domestic workload, ideals of motherhood and stigmas associated to mental health, however, the design of strategies aimed at that purpose requires aditional and intersectoral investigation.
In addition, it is worth mentioning that the variables included in the risk profile (score in the items 2 of EPDS -anhedonia-and 5 -panic and fear-and history of previous treatment of depression) found support in the literature, which reports that access to treatment for depression is associated with depressive symptom levels (or degree of disability) and history of treatment of the disease (14)(15)(16)20) . For example, self reporting is generally considered reliable as antecedent of previous treatment of depression, in cases in which this information is not registered in the medical records (23) . In the case of the scoring obtained by mothers in the items 2 and 5 of EPDS, it is worth mentioning that nursing professionals perform an universal screening using this instrument at postpartum monitoring of child health, at which time it is investigated the suspected of PPD, representing an opportunity to access treatment for the disease (10) .
Therefore This involves to properly inform the mothers about their possible depression, motivate them to adhere to treatment and prioritize the availability of hours for care (11,24) .

Conclusion
In conclusion, it is considered that this study opens up a wide field for further research aiming at the establishment of a risk profile for the lack of access to treatment for women with PPD in PHC. This is a pragmatic predictive model that could guide the human resources available at PHC, to support the implementation of activities aimed at filling the gaps in the treatment of a disease, which has been recognized as a public health problem. In the same vein, it is suggested that nurses be attentive to those mothers with PPD that have low anhedonia, or lack thereof, without panic or fear and no history of depression, since these are the patients who are more likely to not start the treatment for the disease, according to the model. Further studies are needed to validate and evaluate the impact of using this risk profile in real clinical settings.