Locally contextualizing understandings of depression, the EPDS, and PHQ-9 among a sample of postpartum women living with HIV in Malawi

Conversational assessment using artificial intelligence is as clinically useful as depression scales and preferred by users

BACKGROUND

Depression is prevalent, chronic, and burdensome. Due to limited screening access, depression often remains undiagnosed. Artificial intelligence (AI) models based on spoken responses to interview questions may offer an effective, efficient alternative to other screening methods.

OBJECTIVE

The primary aim was to use a demographically diverse sample to validate an AI model, previously trained on human-administered interviews, on novel bot-administered interviews, and to check for algorithmic biases related to age, sex, race, and ethnicity.

METHODS

Using the Aiberry app, adults recruited via social media (N = 393) completed a brief bot-administered interview and a depression self-report form. An AI model was used to predict form scores based on interview responses alone. For all meaningful discrepancies between model inference and form score, clinicians performed a masked review to determine which one they preferred.

RESULTS

There was strong concurrent validity between the model predictions and raw self-report scores (r = 0.73, MAE = 3.3). 90 % of AI predictions either agreed with self-report or with clinical expert opinion when AI contradicted self-report. There was no differential model performance across age, sex, race, or ethnicity.

LIMITATIONS

Limitations include access restrictions (English-speaking ability and access to smartphone or computer with broadband internet) and potential self-selection of participants more favorably predisposed toward AI technology.

CONCLUSION

The Aiberry model made accurate predictions of depression severity based on remotely collected spoken responses to a bot-administered interview. This study shows promising results for the use of AI as a mental health screening tool on par with self-report measures.

Investigating the PHQ-9 With Mokken Scale Analysis and Cognitive Interviews

Scores on the Patient Health Questionnaire-9 (PHQ-9) are frequently used to assess depression both in research and in clinical practice. The aim was to examine the validity of the PHQ-9 sum score by using Mokken scale analysis (Study I) and cognitive interviews (Study II) on the Icelandic version of PHQ-9. A primary care sample of 618 individuals was used in Study I. The results indicate that the PHQ-9 items are not close enough to perfectly unidimensional for their sum score to accurately order people on the depression severity dimension. In Study II, the sample consisted of 53 individuals, with 28 having a history of depression and 25 not. The findings reveal a number of issues concerning respondents' use of the PHQ-9. No systematic differences were found in the results of the two groups. The PHQ-9 sum score should thus be interpreted and used with great care. We provide scale revision recommendations to improve the quality of PHQ-9.

Prevalence of early postpartum depression and associated risk factors among selected women in southern Malawi: a nested observational study

BACKGROUND

The birth of a child should be a time of celebration. However, for many women, childbirth represents a time of great vulnerability to becoming mentally unwell, a neglected maternal morbidity. This study aimed to determine the prevalence of early postpartum depression (PPD) and its associated risk factors among women giving birth at health facilities in southern Malawi. Identifying women vulnerable to PPD will help clinicians provide appropriately targeted interventions before discharge from the maternity ward.

METHOD

We conducted a nested cross-sectional study. Women were screened for early PPD using a locally validated Edinburgh Postpartum Depression Scale (EPDS) as they were discharged from the maternity ward. The prevalence of moderate or severe (EPDS ≥ 6) and severe (EPDS ≥ 9) PPD was determined, including 95% confidence intervals (CI). Data on maternal age, education and marital status, income source, religion, gravidity, and HIV status, among others, were collected during the second trimester of pregnancy, and obstetric and infant characteristics during childbirth were examined as potential risk factors for early PPD using univariable and multivariable logistic regression analyses.

RESULTS

Data contributed by 636 women were analysed. Of these women, 9.6% (95% CI; 7.4-12.1%) had moderate to severe early PPD using an EPDS cut-off of ≥ 6, and 3.3% (95% CI; 2.1-5.0%) had severe early PPD using an EPDS cut-off of ≥ 9. Multivariable analyses indicated that maternal anaemia at birth (aOR; 2.65, CI; 1.49-4.71, p-value; 0.001) was associated with increased risk for moderate and/or severe early PPD, while live birth outcome (aOR; 0.15, 95% CI; 0.04-0.54, p-value; 0.004), being single compared to divorced/widowed (aOR; 0.09, 95% CI; 0.02-0.55, p-value; 0.009), and lower education level (aOR; 0.36, 95% CI; 0.20-0.65, p-value; 0.001) were associated with decreased risk. Being HIV positive (aOR; 2.88, 95% CI; 1.08-7.67, p-value; 0.035) was associated with severe PPD only.

CONCLUSION

The prevalence of early PPD was slightly lower in our selected sample compared to previous reports in Malawi and was associated with maternal anaemia at birth, non-live birth, being divorced/widowed and HIV-positive status. Therefore, health workers should screen for depressive symptoms in women who are at increased risk as they are discharged from the maternity ward for early identification and treatment.

Mental health and life-course shocks in a low-income country: Evidence from Malawi

Economic insecurity has been widely hypothesized to be an important determinant of mental health, but this relationship has not been well-documented in low-income countries. Using data from the Mature Adults Cohort of the Malawi Longitudinal Study of Families and Health (MLSFH-MAC), we investigate the association of negative economic shocks with mental health outcomes such as depression and anxiety among adults aged 45+ years living in a low-income country. Using fixed effects estimates that control for time-invariant unobserved individual heterogeneity, we find that increased economic instability caused by events such as death of a family member, yield loss, or income loss is positively associated with worse mental health outcomes as measured by the PHQ-9 and GAD-7 instruments. Our results suggest that costly economic events are a key component to worsening mental health in settings characterized by pervasive poverty and underscore the importance of mental health as a public health and development target.

Probable perinatal depression and social support among women enrolled in Malawi's Option B+ Program: A longitudinal analysis

BACKGROUND

Malawi's PMTCT Option B+ program has expanded the reach of ART services among pregnant and breastfeeding women, but retention in lifelong HIV care remains challenging. Given that depression can undermine retention, it is important to understand how depression changes over the perinatal period, varies across treatment and retention groups, and could be buffered by social support.

METHODS

Data are from an observational study conducted among women enrolled in Malawi's PMTCT Option B+ program. We used multilevel generalized linear models to estimate the odds of probable depression by time, treatment and retention group, and social support. Probable depression was assessed with the Edinburgh Postnatal Depression Scale and Patient Health Questionnaire-9.

RESULTS

Of 468 women, 15% reported probable depression at antenatal enrollment and prevalence differed across newly diagnosed individuals, second line therapy users, and previous defaulters (18%, 21%, 5%, p = 0.001). Odds of probable perinatal depression decreased over time (OR per month: 0.87, 95% CI: 0.82-0.92) but were higher among those newly diagnosed (OR: 3.25, 95% CI: 1.59-6.65) and on second line therapy (OR: 3.39, 95% CI: 1.44-7.99) as compared to previous defaulters. Odds of probable postpartum depression were lower for participants with high social support (OR: 0.19, 95% CI: 0.09-0.39).

LIMITATIONS

Lack of diagnostic psychiatric evaluation precludes actual diagnosis of depression.

CONCLUSIONS

Probable depression varied across the perinatal period and across treatment and retention groups. Social support was protective for postpartum depression among all participants. Depression screening and provision of social support should be considered in PMTCT programs.

Is Validating the Cutoff Score on Perinatal Mental Health Mood Screening Instruments, for Women and Men from Different Cultures or Languages, Really Necessary?

Background: The most commonly used mood screening instrument in perinatal health is the Edinburgh Depression Scale. The screen-positive cut-off score on this scale, as for others, has been determined, via validation techniques, for over 20 languages/cultures, and for both women and men. While such validation appears to be considered essential, there are studies that could be interpreted to suggest that this is not an important consideration. Methods: Selective studies have been chosen to indicate these opposing points of view. Results: Examples of studies that support the notion of validating cut-off scores are described, as are examples of studies that appear not to support this point of view. Conclusions: (i) Clinical services and researchers need to be mindful of these opposing points of view, and openly discuss them when using screening cut-off scores for their respective populations. (ii) Researchers and Journals need to be more rigorous in ensuring this issue is correctly reported in studies, and/or openly discussed when relevant.

Facilitators, barriers and potential solutions to the integration of depression and non-communicable diseases (NCDs) care in Malawi: a qualitative study with service providers

Background

Integration of depression services into infectious disease care is feasible, acceptable, and effective in sub-Saharan African settings. However, while the region shifts focus to include chronic diseases, additional information is required to integrate depression services into chronic disease settings. We assessed service providers’ views on the concept of integrating depression care into non-communicable diseases’ (NCD) clinics in Malawi. The aim of this analysis was to better understand barriers, facilitators, and solutions to integrating depression into NCD services.

Methods

Between June and August 2018, we conducted nineteen in-depth interviews with providers. Providers were recruited from 10 public hospitals located within the central region of Malawi (i.e., 2 per clinic, with the exception of one clinic where only one provider was interviewed because of scheduling challenges). Using a semi structured interview guide, we asked participants questions related to their understanding of depression and its management at their clinic. We used thematic analysis allowing for both inductive and deductive approach. Themes that emerged related to facilitators, barriers and suggested solutions to integrate depression assessment and care into NCD clinics. We used CFIR constructs to categorize the facilitators and barriers.

Results

Almost all providers knew what depression is and its associated signs and symptoms. Almost all facilities had an NCD-dedicated room and reported that integrating depression into NCD care was feasible. Facilitators of service integration included readiness to integrate services by the NCD providers, availability of antidepressants at the clinic. Barriers to service integration included limited knowledge and lack of training regarding depression care, inadequacy of both human and material resources, high workload experienced by the providers and lack of physical space for some depression services especially counseling. Suggested solutions were training of NCD staff on depression assessment and care, engaging hospital leaders to create an NCD and depression care integration policy, integrating depression information into existing documents, increasing staff, and reorganizing clinic flow.

Conclusion

Findings of this study suggest a need for innovative implementation science solutions such as reorganizing clinic flow to increase the quality and duration of the patient-provider interaction, as well as ongoing trainings and supervisions to increase clinical knowledge.

Trial registration This study reports finding of part of the formative phase of “The Sub-Saharan Africa Regional Partnership (SHARP) for Mental Health Capacity Building—A Clinic-Randomized Trial of Strategies to Integrate Depression Care in Malawi” registered as NCT03711786