Risk groups in children under six months of age using self-organizing maps.

Area of Science:

• Pediatrics
• Public Health
• Data Science

Background:

• Fetal and infant growth is irregular, especially in developing nations, influenced by living conditions and pregnancy complications.
• Identifying distinct risk profiles is crucial for targeted interventions in early childhood development.
• Rio de Janeiro provided the setting to examine growth patterns in a diverse urban population.

Purpose of the Study:

• To identify distinct risk profiles for growth development among infants under six months.
• To apply cluster analysis for grouping infants based on health and birth characteristics.
• To inform public health policies for improved infant outcomes.

Main Methods:

• Employed unsupervised learning, specifically self-organizing maps (SOM) and K-means clustering.
• Analyzed data from 904 infants (496 girls, 508 boys) from 27 primary health care units.
• Utilized maternal interviews and child health card reviews for data collection.

Main Results:

• Four distinct infant groups emerged based on growth and health indicators.
• Group 1: Cesarean births, multiparous mothers, high prematurity/low birthweight, high exclusive breastfeeding, low hospitalization.
• Group 2: Best perinatal/infant health indicators.
• Group 3: Worst perinatal/infant health indicators, requiring urgent intervention.
• Group 4: Higher birthweight, low exclusive breastfeeding, higher hospitalization rates.

Conclusions:

• Self-organizing maps effectively identified infant groups with shared characteristics.
• Infants in group 3 present the highest risk, necessitating immediate public health attention and policy adjustments.
• While groups were identified, no significant differences in postnatal weight adequacy were observed across all groups.