This title appears in the Scientific Report :
2020
Please use the identifier:
http://hdl.handle.net/2128/26228 in citations.
Accurate early identification of postpartum depression using demographic, clinical and digital phenotyping
Accurate early identification of postpartum depression using demographic, clinical and digital phenotyping
Introduction: Postpartum depression (PPD) affects up to 13% of women. Although demographic and clinical risk factors have been identified, there are no accurate predictors for PPD to such an extent that at risk mothers can be identified and benefit from early interventions.Methods: We recruited 308...
Saved in:
Personal Name(s): | Hahn, Lisa (Corresponding author) |
---|---|
Eickhoff, Simon / Habel, Ute / Stickeler, Elmar / Goecke, Tamme W. / Stickel, Susanne / Schnakenberg, Patricia / Franz, Matthias / Dukart, Jürgen (Last author) / Chechko, Natalia (Last author) | |
Contributing Institute: |
Gehirn & Verhalten; INM-7 |
Imprint: |
2020
|
Conference: | 2020 Society of Biological Psychiatry Annual Meeting, New York (virtual) (USA), 2020-04-30 - 2020-05-02 |
Document Type: |
Poster |
Research Program: |
Theory, modelling and simulation |
Link: |
OpenAccess |
Publikationsportal JuSER |
Introduction: Postpartum depression (PPD) affects up to 13% of women. Although demographic and clinical risk factors have been identified, there are no accurate predictors for PPD to such an extent that at risk mothers can be identified and benefit from early interventions.Methods: We recruited 308 mothers at the University Hospital Aachen. Demographic and clinical measures incl. self-reported mood and stress assessment scales were collected from two to five days to 12 weeks postpartum, on a weekly and daily basis respectively. After 12 weeks, participants were defined into three groups according to DSM-5 criteria: healthy controls (HC; N=247), women with PPD (N=28), and women with adjustment disorder (AD; N=33). We used a logistic regression algorithm to evaluate the potential predictive power of baseline demographic, clinical, and digital phenotyping for early identification of PPD. We performed 1000 permutations using three-fold cross validation to obtain accuracy estimates.Results: Most accurate early differentiation between PPD vs. HC and AD vs. HC was achieved by using baseline demographic and clinical risk factors in addition to postnatal depression scores at week 3 (PPD vs. HC: balanced accuracy: 0.78, sensitivity: 0.73, specificity: 0.82; AD vs. HC: balanced accuracy: 0.89, sensitivity: 0.85, specificity: 0.93). Accurate differentiation of PPD vs. AD was only possible at week 6 with mood scores being most accurate resulting in a balanced accuracy of 0.76 (sensitivity: 0.76, specificity: 0.76).Conclusion: In conclusion, combinations of mood level, postnatal depression scores, and baseline risk factors allowed for an accurate early identification of women at risk for PPD. |