Validity of the Language Environment Analysis in two neonatal intensive care units supporting parent-infant closeness

• In Event: Poster Session 02
  In Poster Session: PS 02 Topic - Language, Communication

Abstract

Background and aims: preterm children are at risk for weak language skills (Barre et al., 2011), but the language development can be affected by various factors, including early environment (DeMaster et al., 2019). The infant’s natural sound and language environment can be recorded and analyzed using the Language Environment Analysis (LENA) system (LENA Research Foundation). The LENA system has been used in a few studies to study the language environment in the neonatal intensive care unit (NICU) and early vocalizations of preterm infants. Findings suggest a connection between the amount of parental talk at 32 weeks’ gestational age and the number of preterm infants’ vocalizations at this age point (Caskey et al., 2011), and the infants’ language development at 7 and 8 months’ corrected age (Caskey et al., 2014). Since the validity of the LENA system in the NICU has not previously been evaluated, it is important to analyze the LENA measures from this context. In addition, features of individual languages can influence the accuracy of LENA-provided measures from different language environments (Ganek et al., 2018). The aim of the present study was to evaluate the validity of the LENA system’s measures in two NICUs supporting parental closeness, in the Finnish and Estonian language settings.

Methods: The sound and language environment of 14 very preterm (born <32 gestational weeks) infants, 7 infants from both units, was recorded at the gestational age of 32 (+2) weeks for 16 hours with the LENA system in the NICUs. At the recording days, parental closeness diaries (Raiskila et al., 2017) were maintained to document parental presence. The analyzed samples, two five-minute chunks with a high percentage of adult speech from every recording, when a parent was present, were selected based on information derived from LENA recordings and from the parental diaries. Agreements between LENA- and human coder-provided values for the following estimates were calculated; female words, male words, adult words, child vocalizations, conversational turns and duration of silence.

Results: Results showed high variations in measured estimates and agreements. In the analyzed samples, the mean number of female words based on LENA estimates was 609 and based on human transcription 698. Furthermore, the mean LENA-provided value of child vocalizations was 14 and the human-provided value was 6. The highest total agreements were found in female words (r=.91, p<0.001) and adult words (r=.95, p<0.001). The agreements for male words, child vocalizations, conversational turns and duration of silence were modest or low (r=-.03—0.64).

Conclusions: Our study provides new validity information on the LENA system in two NICUs supporting parent-infant closeness, and in Finnish and Estonian language environments. Findings show that the LENA-provided estimates for adult words were valid in these NICU settings, especially for female words. This information can be applied when studying parental speech in the NICU, early interventions and language outcomes. LENA-provided estimates for child vocalizations were less valid in this study, compared with estimates based on the human ear. More knowledge is needed about automatic analysis of infant vocalizations in the NICU.

Authors

• Eva Ståhlberg-Forsén, university of helsinki
  Presenting Author

• Reija Latva, Tampere University Hospital
  Non-Presenting Author

• Suvi Stolt, University of Helsinki
  Non-Presenting Author