TALKING ABOUT TO SUPPORT SCREENING IN POSTPARTUM DEPRESSION

L. Deledda;R. Pintus;Dessì Angelica;A. Lai;Maurizio D’Alterio;P. Ranzi;F. Marinaro;G. Oliverio;L. Gaetano;A. Colangelo;S. Angioni;V. Fanos

2022-01-01

---

Abstract

Postpartum depression (PPD) is a mood disorder that affects up to 1/5 of new mothers. It manifests itself with the typical symptoms of major depression and typically occurs within 6 months of delivery. The consequences of this pathology affect both the health of the mother and that of the child, which could undergo alteration of the affective, behavioral and social development, but also on the relationship with the partner (including the paternal postpartum depression). An early diagnosis that allows for the timely establishment of an adequate therapy is therefore essential. Talking About (TA) is a Speech Emotion Recognition project (SER) made by G.P.I. (Trento, Italy) which consists of a series of artificial intelligence algorithms that have the purpose of reliably recognizing emotions and/or stress through audio streams of the human voice (e.g. telephone calls, telemedicine). Voice acquisition takes place by recording an audio while the mother describes an image chosen from a series. The study aims to compare the results obtained through TA and the score obtained with the Edinburgh Postnatal Depression Scale (EPDS) which is the most widely used tool for screening for PPD, in order to evaluate the possible correspondence of the results. Both tests (EPDS and TA) are administered 1-2 months after delivery through the Haumea platform, created by G.P.I. specifically for this purpose, which allows the administration of the tests both in presence, through the use of a tablet, and remotely via video call. In total, 64 women were recruited: 53 in the puerperium, close to childbirth, and 11 by telephone. Consequently, postnatal recruitment proved to be slightly more effective. To date, 30 of these 64 patients have performed the test (46.8%). Only 4 of these 30 patients (13.3%) were able to carry out the test in presence, the remaining 26 (86.7%) chose, for convenience, to carry out the remote test via video call. The test was on average 52 days after delivery. The administration of the test through Haumea is divided into 3 phases: 1. collection of statistical data (age of the mother, gestational age, weight of the newborn at birth, etc.) and more specific data relating to any risk factors for PPD (family history of depression, complications in the peripartum, health problems of the newborn, etc.) in order to stratify the cases examined; 2. EPDS administration; 3. administration of the TA test. A preliminary analysis of the data obtained so far shows how TA is able to provide a specific profile for each new mother. These profiles are presented as a specific combination of 5 different categories that describe the emotional state of each patient: “I feel good”, “It is a good day”, “I do not agree”, “I am tired”, “I am pessimistic”. Specifically, each of these categories is expressed as a percentage, thus highlighting different aspects of the same mother. Examples of profiles obtained are presented hereafter. Example 1: “I feel good” 24.10%; “It is a good day” 32.3%; “I do not agree” 40.8%; “I am tired” 0.1%; “I am pessimistic” 2.7%. Example 2 (patient who did not sleep the whole night): “I feel good” 9%; “It is a good day” 0%; “I do not agree” 0.7%; “I am tired” 90.3%; “I am pessimistic” 0%. Thus, it is interesting to note how with TA it is possible to highlight different emotional facets of a subject through a single test, unlike the single score obtained with the EPDS which inevitably, being a screening tool, is based on specific cut-offs useful for stratifying the risk of PPD. In fact, from the administered tests, it emerges that women who obtained the same score on the EPDS presented more or less different profiles at the TA, precisely because this test investigates different aspects compared to the EPDS. Therefore, it remains to be investigated whether, once the results obtained with TA have been standardized, there is a good overlap of the different profiles belonging to patients who have scored the same score on the EPDS and a correlation between the 2 tests can therefore be identified. This could be of interest in order to further improve the screening process and hypothetically allow for an ever earlier diagnosis of PPD.