Expressive vocabulary development in children with moderate hearing loss – the impact of auditory variables and early consonant production

ABSTRACT

In this study, the early expressive vocabulary development was investigated in a group of children with moderate hearing loss (HL). Size and development of expressive vocabulary from 18 30 months were analyzed and compared to a group of children with normal hearing (NH). For the children with HL, the impact of auditory variables on number of words were examined. The relationship of early consonant production to number of words produced of both groups were examined and the phonological complexity of reported words was compared between the groups. The results showed that children with HL (n = 8) produced a similar number of words as the NH (n = 8) at 18 months, but fewer at 24 and 30 months. Hours of HA use showed significant correlations to number of words. The number of different true consonants at 18 months for the whole group showed a significant relationship to number of words produced at 24 months. No significant differences were found between children with HL and NH children regarding phonological complexity of reported words. The findings indicate that the children born with moderate HL who were fitted with hearing aids (HAs) before 6 months of age are at risk in their development of expressive vocabulary. Full-time use of HAs and monitoring of early consonant use should be encouraged in the early intervention of this target group.

KEYWORDS:

• hearing loss
• expressive vocabulary
• consonant development

Introduction

The acquisition of spoken words is based on the ability to hear speech during infancy (Kuhl, 2004). For children born with hearing loss (HL), acquiring spoken words is challenging due to restrictions in their auditory access (Moeller & Tomblin, 2015). Early detection through universal newborn hearing screening (UNHS) and diagnosis of HL followed by intervention with hearing aids (HAs) have led to improved expressive vocabulary outcomes in children with HL (Ching, 2015). In spite of these improvements, studies of expressive vocabulary development the first three years in children born with HL show that they are still delayed in comparison to children with normal hearing (NH) (Vohr et al., 2011; Yoshinaga-Itano et al., 2017). As vocabulary size has been found to be a strong predictor for future success with literacy and academics (Fenson et al., 1993; Whitehouse et al., 2011), studies investigating the expressive vocabulary development in children with HL and the factors that impact their lower outcomes is motivated.

In typical expressive vocabulary development, the first word appears around 12 months of age. These words predominantly consist of nouns with simple syllable structures, such as CVCV structure (Vihman, 2014). There is an acceleration of growth following the appearance of the first 50 words. This rapid growth occurs around 18–20 months years of age, often referred to as the “vocabulary spurt” (Fenson et al., 1993; Stoel-Gammon, 2011). All children do not demonstrate this period of rapid acceleration (D’Odorico et al., 2001) and there is large individual variability (Berglund & Eriksson, 2000a; Pine et al., 1996). In terms of the large variability, the average number of words produced in individual children by 2 years of age ranges from less than 300 words to more than 500 words (Stoel-Gammon, 2011). Despite the large variability, a child´s vocabulary size is a strong predictor of future language ability (Lee, 2011). Therefore, monitoring the number of words that children with HL acquire over time can be useful as they are a group identified as being at-risk for spoken language difficulties (Ching, 2015; Moeller & Tomblin, 2015).

Research of expressive vocabulary development demonstrates that children with HL are delayed compared to children with NH (Mayne et al., 2000; Moeller et al., 2007a; Vohr et al., 2011). Vohr et al. (2011) found significantly fewer words produced by 30 children with varying degrees of HL at 12–16 months compared to a control group of children with NH. Although the number of words of the children with HL increased with age, they still produced significantly fewer words than children with NH at 18–24 months. Furthermore, the number of children with HL who performed below the 10^th percentile on expressive vocabulary increased with age, demonstrating persisting challenges in the development of expressive language in children with HL. The lower number of words was still present even after controlling for socioeconomic status and neurodevelopmental comorbidity (Vohr et al., 2011).

In a study by Mayne et al. (2000) of 202 children with varying degrees of HL who were followed from 8 to 37 months of age, the average performers (50^th percentile on the expressive vocabulary measure) were significantly delayed in comparison to children with NH. Despite early detection of HL, an acceleration of the expressive vocabulary was not seen until after the age of 25 months which was about six months later compared to the children with NH. Similar findings have been found by Moeller et al. (2007b) where a group of 11 children with mild to severe HL expanded their expressive vocabulary from 10 to 24 months of age but at a much slower rate compared to the group of children with NH. Authors highlight that slow word learning in children with HL is strongly linked to their difficulties with speech perception and is intertwined with their difficulties development of expressive phonology (Moeller et al., 2007b)

In terms of speech perception, young children with NH under 4 years of age have shown significantly poorer speech perception skills in noisy situations compared to quiet (Schafer et al., 2012). In addition, children with HL between the ages of 18–36 months have shown lower scores on functional performance in noisy situations compared to children with NH of the same age (Ching et al., 2013; Persson et al., 2020). Young children spend many hours in preschool, which is an important environment for spoken language stimulation. The fact that preschools have high levels of background noise (Persson Waye et al., 2019) constitute challenges for children with HL to perceive spoken language and consequently, learn new words. In addition to this, spoken language communicated at a group level is often given at a distance, allowing for background noise and reverberation to contaminate the target signal (Sjödin et al., 2012).

The other aspect to consider in expressive vocabulary development is phonology (Stoel-Gammon, 2011). The development of oral stops with anterior placement and the so-called true consonants (excluding glides and glottals) predicts future expressive vocabulary outcomes (Stoel-Gammon, 2011). Studies of children with profound HL show severe delays in the development of early consonant production (Stoel-Gammon, 1988), meanwhile studies of children with moderate HL show mixed results (Löfkvist et al., 2019; Moeller et al., 2007b). What has been found is that canonical babbling in children with HL includes less or even no productions of true consonants which could lead to a lower expressive vocabulary (Moeller et al., 2007a; Persson et al., under review; Stoel-Gammon, 1988). However, studies of the link between early consonant production and expressive vocabulary in this target group is sparse, particularly in children fitted with HAs before the age of 6 months (Löfkvist et al., 2019).

Research of children with HL have found several factors to be associated with their expressive vocabulary development. In a study of children with HL between 8 and 39 months of age it was found that the number of produced words were significantly higher for the children with mild to moderate HL, no additional disabilities and mothers with higher levels of education (Yoshinaga-Itano et al., 2017). Considering that these background factors only explained 41% of the variance in vocabulary outcomes, the authors highlighted the importance of including additional factors of access to the auditory signal that may contribute to this development in future research. One of the suggested factors was adherence with amplification use (Yoshinaga-Itano et al., 2017). Studies of HA use in infants and toddlers with HL have mainly focused on investigating the amount of HA use but have not investigated the possible association with early expressive language. As hours of HA use in this target group has been found to vary greatly (Persson et al., 2020; Walker et al., 2013) the possible link between HA use and expressive vocabulary motivates further investigation.

Another factor that may impact a child with HL from receiving optimal auditory input, and consequently acquire new words, is aided audibility of the HAs. Poorly fitted HAs have been associated with poorer speech and language outcomes in children aged 3–5 years old (Tomblin et al., 2014). Similar to investigations of HA use, aided audibility has been found to be less than optimal in children with HL under 5 years of age (McCreery et al., 2015b), albeit the impact on early speech development of children under 3 years of age is still unclear and in need of further research (Moeller & Tomblin, 2015).

Early fitting and consistent use of optimally fitted HAs provide children with HL a possibility of accessing sounds and spoken language in their environment. With regards to restrictions in auditory access (e.g. low functional performance in noise, low HA use, non-optimal aided audibility) and limitations in early consonant production (e.g. few productions of true consonants in babbling and early speech), it could be further hypothesized that children with HL are less likely to perceive words that are phonologically complex compared to children with NH. To date, studies on expressive vocabulary in children with HL have focused on vocabulary size (Pittman et al., 2005) but fewer studies have examined phonological complexity of expressive vocabulary (Välimaa et al., 2018).

To complement previous studies on vocabulary development of children with HL, the purpose of this study was to investigate the longitudinal impact of auditory access nd early consonant production on number of words produced in a contemporary group of children with moderate HL who were all fitted with HAs before the age of 6 months. The following research questions were posed:

• What is the size and growth of the expressive vocabulary in children with moderate HL at 18–30 months compared to children with NH?

• What is the relationship between the number of different true consonants produced by the children with HL and NH at 18 months to the number of words produced at 18, 24 and 30 months?

• Does age at amplification, amount of HA use, aided audibility, auditory development and functional auditory performance impact the number of words produced in the children with HL from 18 to 30 months of age?

• Do children with moderate HL between the ages of 18–30 months learn words that are less phonologically complex compared to children with NH of the same age?

Methods

Participants

Two groups of children were included in this study. A study group of children with HL were recruited from the Department of Hearing Habilitation at Karolinska University Hospital, Stockholm (HHKUS), Sweden, between 2015 and 2016. The HHKUS has an uptake of children born with HL in the whole Stockholm region, with around 2.3 million inhabitants. Children with a diagnosed sensorineural bilateral HL (ICD-code H90.3) of mild to moderate degree (four-frequency pure-tone average/4 FPTA) of 30–60 dB HL) who received HAs before the age of 6 months were invited to take part in the study. Inclusion criteria were that one of the parents needed to be a native speaker of Swedish, no diagnosed syndrome at recruitment, and data at all time points. During the recruitment period, parents of 15 children with HL were eligible for inclusion of which 14 agreed to participate. All children had been fit with HAs on both ears between the ages of 2–6 months. During the study, six children were excluded due to improved hearing (n = 1), progressive hearing loss resulting in cochlear implantation (n = 2), suspected comorbidity (n = 1), relocation after the first data collection (n = 1), and missing data (n = 1), resulting in eight participants. The etiology of the remaining children HL was unknown in seven children and one was hereditary. The parents of all children with HL were native Swedish speakers and all children went to mainstream preschools (all had started by the age of 18 months). Descriptive data of age at diagnosis, age at amplification, mean best ear four-frequency pure-tone average (4 F PTA), and other auditory variables of each of the eight participants with HL can be found in Table 1.

Table 1. Descriptive data of the children with HL. Unaided four-frequency (4 F) pure-tone average (PTA): (500, 1000, 2000 and 4000 Hz), mean dB HL over time measured at 10, 18, 24 and 30 months; age at amplification; hours of hearing aid (HA) is from datalogging of the HAs at each of the ages. Aided audibility is presented as Speech Intelligibility Index (SII values). Auditory development is from results on the LittlEARS Auditory Questionnaire®/LEAQ and scores of functional auditory performance in noise with Parent´s Evaluation of Aural/Oral Performance in Children /PEACH.

Participant	P1	P2	P3	P4	P5	P6	P7	P8	Mean (SD)
Unaided 4F PTA (dB HL) (mean over time) Right/Left	45.1/44.8	48.6/48.1	44.7/44.6	46.0/37.3	43.5/43.5	41.0/40.0	58.8/59.0	45.5/44.8	45.3 (6.5) Right 46.6 (5.1) Left
Age at diagnosis (months)	2	2	1	3	3	2	2	4	2.4 (0.9)
Age at amplification (bilateral fitting) (months)	3	4	3	5	4	5	6	5	4.6 (1.2)
HA use (hours) 10 mo 18 mo 24 mo 30 mo	8.4 7.8 9 9	8.4 8.5 11.1 11.3	4 3 7.6 8.8	7.5 4 7.1 9.7	6.4 9 11.5 6.8	11.4 11.3 10.1 10.2	9.0* 8.6 9.2 10.6	9.4 10.2 9.2 11	7.5 (2.5) 7.8 (2.9) 9.3 (1.5) 9.7 (1.5)
Aided audibility (SII) 10 mo 18 mo 24 mo 30 mo	.69 .78 .90 .70	.50 .51 .52 .53	.76 .81 .83 .66	.72 .53 .83 .86	.49 .73 .65 .73	.68 .72 .78 .46	.41 .53 .64 .71	.49 .78 .77 .79	0.59 (.13) 0.67 (.15) 0.71 (.14) 0.65 (.13)
Auditory development (LEAQ score) 10 mo 18 mo 24 mo	20 31 35	18 29 34	21 29 33	21 33 33	17 30 34	17 31 35	7 25 28	21 31 35	18 (5) 30 (2) 33 (2)
Functional auditory performance (PEACH score in noise) 18 mo 24 mo 30 mo	15 16 15	14 15 17	12 15 15	19 17 16	9 9 12	12 17 13	10 14 13	14 16 13	13.1 (3.1) 14.8 (2.6) 14.2 (1.7)

Note: *Based on parental report as the datalogging was based on less than 1 week of HA use due to change in HAs.

A reference group of eight children with NH was recruited from two child health care centers in the same region as the children with HL. The same inclusion criteria as for the children with HL were held for the reference group, except the reference group needed to have NH at birth (controlled by UNHS).

As level of parental education has been found to impact expressive vocabulary scores (Fenson et al., 1993; Rowe et al., 2016), all parents were asked to complete their level of education on a three-point scale (college = 1, high school = 2, university = 3). Across both groups, 75% of the mothers and 55% of the fathers reported to have a university degree. This is higher in relation to the mean at a national level where 50% of the women and 38% of the males have a university degree (Statistics Sweden, 2019). The summarized level of education of both parents (median, range) was similar between the groups: children with HL (6, 3–6); children with NH (6, 4–6), and the difference was not statistically significant: U (HL = 8, NH = 8) = 28.5, Z = -.424, p = .67.

To control for normal cognitive development, all participants were evaluated with the Bayley Scales of Development (Bayley-III) (Bayley, 2006) at 3 years of age. The Bayley-III testing was performed by a clinical psychologist. One child with HL did not complete the assessment due to behavioral issues. All other children (children with HL, n = 7, children with NH, n = 8) presented with scores within in the normal range. The scores on the cognitive scale (median, range) were similar between groups; HL (10, 7–12), NH (10, 8–14), and the difference was not statistically significant: U (HL = 7, NH = 8) = 20.5, Z = -.919, p = .36.

Materials and procedure

Ethical approval for the current study was obtained from the Regional Ethical Committee in Stockholm (Dnr.2014/1162-31/1). All participating families were given verbal and written information about the study before enrollment and all parents provided written consent.

Assessment of hearing status

For the children with HL, an otolaryngologist confirmed the diagnosis (congenital sensorineural bilateral HL, ICD-code H90.3) before study enrollment. Children with HL had a confirmed diagnosis of moderate HL through a battery of electrophysiological tests. Behavioral audiograms at 10, 18, 24 and 30 months of age utilized observation audiometry, visual reinforcement audiometry, or conditioned play audiometry to obtain air and bone thresholds. Ear specific behavioral air and bone conduction thresholds were obtained at 500, 1000, 2000 and 4000 Hz (MADSEN Astera2, GN Otometrics A/S Denmark) with insert earphones (Otometrics Oto insert ER-3A from Etymotic Research Inc.) and bone oscillator (Radioear B71). Seven of the children were initially fit with Widex Baby 440, of which four changed to Oticon Sensei Pro 75 during the course of the study. One child was initially fit with Oticon Sensei Pro 75 and did not change HAs.

The hearing of the children with NH was screened (American National Standards Institute, 2010) at 20 dB HL at 500, 1000, 2000 and 4000 Hz (MADSEN Astera2, GN Otometrics A/S Denmark) by a pediatric audiologist at 10, 18, and 24 months of age. In addition, screening was performed at 36 months of age instead of 30 months because of resource constraints. In cases where insert earphones (Otometrics Oto insert ER-3A from Etymotic Research Inc.) could not be used, the assessment was carried out under headphones (Radioear DD45, USA). Two children at 18 months, and one child at 24 months performed the hearing assessments in soundfield.

Auditory variables

Auditory variables examined included age at amplification, hours of HA use, aided audibility, auditory development and functional auditory performance in noise (see Table 1). Age at amplification was reported by the parents at study start and confirmed by the pediatric audiologist through medical records. A paediatric audiologist read the datalogging of the HAs at the ages of 10, 18, 24 and 30 months to obtain the average number of hours used between datapoints.

To investigate the aided audibility of the HAs, a calculation of Speech Intelligibility Index (SII), Situational Hearing Aid Response Profile software (SHARP 1997, version 7) was utilized. The age of each participant, audiometric thresholds (air and bone thresholds) and HA output levels with an input of 65 dB SPL were entered in SHARP for all participants at 10, 18, 24 and 30 months of age. The SHARP program requires at least four frequencies (500–4 kHz) for SII calculation. The SII ranges from 0 to 1 with zero representing no audibility of the speech spectrum and one representing a full audibility of the speech.

The Swedish version of the LittlEARS Auditory Questionnaire® (LEAQ) was used to collect data on auditory development (Coninx et al., 2009; Persson et al., 2019). The parents completed the LEAQ every other month from study entry until their child turned 2 years of age. The LEAQ consists of 35 Y/N questions to which the parents respond if they have observed a variety of auditory behaviors of their child (including responses to receptive and expressive speech/language).

To collect information around functional auditory performance, the Swedish version of the Parents´ Evaluation of Aural/Oral Performance in Children (PEACH) (Brännström et al., 2014; Ching & Hill, 2007) was used. The PEACH consists of 13 questions about how children respond to sounds in their environment. Parents were asked to observe their child and answer the questions on a 4-point Likert scale. The PEACH includes questions around functional performance in both quiet and noisy situations which are added into a total score. In the current study, the parents of the children with HL completed the PEACH when their children were 18, 24 and 30 months.

Vocabulary size and phonological complexity of parent-reported words

The Swedish version of the MacArthur-Bates Communicative Developmental Inventory (MCDI) – Words and Sentences (SECDI-II) was used to document the expressive vocabulary development (Berglund & Eriksson, 2000b; Fenson et al., 1993). The SECDI-II was completed by the parents when their children were 18, 24, and 30 months old (± 2 weeks). Parents were instructed to tick each word of the SECDI-II they interpreted their child was producing (using in their speech) regardless of the child´s pronunciation of the word.

The phonological complexity of the parent-reported words on the SECDI-II was calculated based on the Swedish version of the Word Complexity Measure (WCM-SE; Marklund et al., 2018a), originally developed for English by Stoel-Gammon (2010). The WCM-SE is a measure that calculates the phonological complexity of words or utterances, based on a number of complexity parameters (Tables 2 and 3). The WCM-SE is usually used to analyze standardized recordings of children. In this study, however, the purpose was not to investigate phonological complexity of the children’s produced words, but to explore if the children with HL were reported to use words that in their “adult” target form have the same complexity as the words reported for children with NH.

Table 2. The complexity parameters in the Swedish version of the Word Complexity Measure (WCM-SE) (Marklund et al., 2018a) and the points given for each parameter.

Complexity parameter			Points
1 2	Word patterns	More than two syllables Stress on non-initial syllable	1 per word 1 per word
3 4	Syllable structure	Word-final consonant Consonant cluster	1 per word 1 per occurrence
5 6 7 8 9 10	Sound classes	Velar consonant Liquid /r/ Long, front, rounded vowel Fricative Voiced fricative	1 per occurrence 1 per occurrence 2* per occurrence 1 per occurrence 1 per occurrence 1 per occurrence (in addition to 1 point for being a fricative)

*Each occurrence of /r/ is consistently given 2 points (equated to a voiced fricative). Variations of /r/ that would give 1 point (flaps) or 3 points (trills) are not taken into account since calculations are made on target words (SECDI items) and not on actual productions.

Table 3. Swedish version of Word Complexity Measure (WCM-SE) calculations of five SECDI-II words with complexity sum ranging from 0 to 8. Columns represent Swedish word, phonetic transcription according to IPA, English translation of the word, points given to occurring WCM-SE complexity parameters, and complexity sum for each word.

Target word			Complexity parameter
Swedish word	Phonetic transcription (IPA)	English translation	1	2	3	4	5	6	7	8	9	10	Sum
tå	[to:]	toe											0
min	[mɪn]	my/ mine			1								1
katt	[kat]	cat			1		1						2
vatten	[ʋatɛn]	water			1						1	1	3
snögubbe	[snø:ɡubɛ]	snowman	1			1	1			1	1		5
lekpark	[lekpark]	playground			1	2	2	1	2				8

Consonant production

A standardized observation of early consonant use (Lieberman & Lohmander, 2014; Lohmander et al., 2017) was audio and video recorded by the first author when the children were 18 months (external microphone Røde NT4, audio recorder TASCAM DR-22WL, video camera Panasonic HC-V750). The video and audio recordings were approximately 30–45 minutes and consisted of parent-child interactions during free play. Parents were instructed to play and interact as they usually do with their child, positioned on a mattress and given a fixed set of age-appropriate toys.

Analysis

Vocabulary size: The total number of words on the SECDI-II (part A, vocabulary checklist including onomatopoetica) was calculated by the first author. Words added next to the referent word by parents (e.g. child´s “own word” or “the word in another language”) were excluded from the analyses. Sixty percent of the material was cross-checked by a linguistic student.

Phonological complexity score of produced words: For the analyses of the phonological complexity score of the parent reported words in SECDI-II the onomatopoetic words were excluded as they are more prone to be irregularly produced resulting in large variations in score (Marklund et al., 2018a). This resulted in a total number of possible words used for calculation in the SECDI-II to be 697, ranging in complexity scores of 0–12 (total complexity sum of all words = 2559, M = 3.6). As the words may have different scores depending on differences in Swedish dialect, the words were analyzed in respect to the dialect of the majority of people in the region where the participants were recruited. Each of the produced words ticked for each child on the SECDI-II were analyzed based on the WCM-SE, and a total complexity sum for all words produced were calculated for each child. The total complexity sum was then divided with the number of words that were reported for each child at each age, resulting in a mean complexity level for words for each child at all ages (Table 4).

Table 4. Descriptive statistics for each participant in the group of children with normal hearing (NH) and children with hearing loss (HL), on number of different true consonants at 18 months, number of words produced based on parent reports (SECDI-II) and (in brackets) mean complexity sum of the words reported at 18, 24 and 30 months.

Participants	Number of different true consonants	Number of reported words and mean phonological complexity of target form of reported words
NH	18 months	18 months	24 months	30 months
ID1	3	2 (1.00)	47 (1.85)	381 (2.48)
ID2	7	6 (0.83)	36 (2.05)	421 (2.36)
ID3	4	17 (1.17)	200 (2.27)	334 (2.35)
ID4	6	26 (1.11)	376 (2.26)	650 (2.49)
ID5	5	149 (2.14)	243 (2.42)	565 (2.45)
ID6	4	13 (1.69)	44 (2.23)	115 (2.37)
ID7	8	30 (1.93)	280 (2.44)	337 (2.41)
ID8	8	11 (1.64)	360 (2.49)	349 (2.50)
HL
P1	0	83 (1.80)	241 (2.27)	398 (2.47)
P2	3	4 (1.50)	140 (2.52)	404 (2.55)
P3	0	15 (1.47)	26 (1.50)	69 (1.75)
P4	2	26 (1.54)	66 (2.17)	152 (2.31)
P5	2	6 (1.50)	34 (1.85)	87 (2.01)
P6	1	10 (1.40)	41 (2.39)	310 (2.40)
P7	1	5 (1.60)	77 (2.51)	277 (2.46)
P8	3	68 (0.79)	177 (2.29)	431 (2.39)

Consonant production: The number of different true consonants were assessed from the audio-video recordings by in total five trained listeners who are speech and language pathologists (SLPs). The consonants perceived in at least two different occasions during the observation of the recording were noted and counted for each participant. Details of the assessment and analysis of early consonant production can be found in Lieberman and Lohmander (2014). For reliability purposes, approximately 25% of the observations were re-assessed by one SLP and in cases of substantial disagreement, a SLP and researcher with extensive experience (third author) performed the assessment.

Statistical analysis

The number of produced words from the SECDI-II at 18, 24 and 30 months are presented in a descriptive manner (raw data and percentiles) for each group. Comparisons between group means and in relation to Swedish reference data (Wordbank, n = 420 boys) at 18, 24 and 30 months of age were made. A two-way repeated measures ANOVA was performed with group as the between subject factor and age as the within-subjects factor to examine the growth in the number of words across ages.

The number of different true consonants at 18 months were examined in relation to the number of produced words at the ages of 18, 24 and 30 months, using Spearman´s rank correlation. Analyses were made for each group and the whole group.

For the children with HL, the impact of the auditory variables (i.e. age at amplification, amount of HA use, aided audibility, auditory development and functional auditory performance) on the number of words on the SECDI-II at 18, 24 and 30 months was examined through Spearman´s rank correlations.

Mean complexity score for reported words at the different ages was calculated for each child.

Results

Size and growth of number of words produced

Large individual variability in number of words produced was found. See Table 4 and Figure 1 for details. The number of words did not differ between the groups at 18 months (Median HL = 20.5, NH = 20.5) but differed at 24 months (Median HL = 79, NH = 272) and at 30 months (Median HL = 302, NH = 375). However, a Mann-Whitney U-test indicated that the differences were not statistically significant at any of the ages: 18 months: U (HL = 8, NH = 8) =30.5, Z = −.158, p = .87, and at 24 and 30 months respectively: U (HL = 8, NH = 8) =19.0, Z = −1.365, p = .17.

Figure 1. Number of produced words on the SECDI-II for children with HL and NH at ages 18, 24 and 30 months (n = 8) in both groups at all ages.

In terms of growth over time, all children increased their number of words from 18 to 30 months with the exception of one child with NH who had a decrease of 11 words between 24 and 30 months (from 367 to 358). A two-way repeated-measures ANOVA of the children with HL and NH, with group as the between subjects factor and age as the within subjects factor, revealed a significant increase in number of expressive words with age, F (1, 74) = 109,679, p = .00 (Figure 2). The children with NH had scores similar to the reference group of children retrieved from the Wordbank (n = 420, boys only). All groups presented with a similar mean number of expressive words at 18 months, but at 24 months the children with HL showed a flatter curve resulting in a gap between them and children with NH and those from the Wordbank. This gap widened even further at 30 months.

Figure 2. Estimated marginals mean number of produced words on the SECDI-II. The dotted line represents the children with HL, the dashed line the children with NH and the solid line is a group of 420 children used as a reference for Swedish children with NH (boys only to match the study participants) from the Wordbank. The children with HL and NH were (n = 8) at all ages.

The number of words of the children with HL and children with NH were around the same percentile levels in relation to the normative data of the SECDI-II at 18 months of age (HL 40^th percentile, NH 44^th). At 24 months the mean percentiles were 30 (HL) and 55 (NH) respectively, and at 30 months 53 (HL) and 78 (NH). In terms of individual variability, more children with HL, compared to the NH, were performing just at, or below, the 10^th percentile at all ages compared to the children with NH.

Relationship between number of different true consonants and number of produced words

The number of different true consonants at 18 months ranged from 0 to 3 in the children with HL (Median 1.5) and 3 to 9 in children with NH (Median 6.5). The correlation between number of words on the SECDI-II at 18, 24 and 30 months for each group and the group as a whole are found in Table 5. The only statistically significant correlation was found for the group as a whole between number of different true consonants at 18 months and number of produced words at 24 months (rho = 0.54, p <.05).

Table 5. Spearman rank correlations (rho) between the total number of different true consonants at 18 months to number of words produced on the SECDI-II at 18, 24 and 30 months. Results are presented for the whole group and each group: children with HL (n = 8) and children with NH (n = 8).

Assessment	Study group	Total number of parent-reported words produced on the SECDI-II (rho)
		18 months	24 months	30 months
Number of different true consonants at 18 months	Whole group	0.13	0.54*	0.47
	HL	−0.20	0.24	0.49
	NH	0.34	0.52	0.02

*Correlation is significant at the 0.05 level

Impact of auditory variables on the number of produced words

Of the auditory variables examined, hours of HA use and scores on the LEAQ showed a significant association on the number of expressive words. See Table 6 for an overview of the correlation analyses on auditory variables to number of words produced.

Table 6. Spearman rank correlations (rho) between auditory variables and total number of words produced on the SECDI-II at 18, 24 and 30 months in the children with hearing loss.

Auditory variable	Number of words 18 months	Number of words 24 months	Number of words 30 months
Age at hearing aid fitting	-0.64	0.25	0.42
Hours of HA use 10 months 18 months 24 months 30 months	0.41 -0.21 N/A N/A	0.54 0.12 -0.02 N/A	0.76* 0.45 0.29 0.79*
Aided audibility (SII) 10 months 18 months 24 months 30 months	0.53 0.48 N/A N/A	-0.28 -0.28 -0.13 N/A	-0.36 -0.53 -0.38 -0.26
Auditory development (LEAQ score) 10 months 18 months 24 months	0.75* 0.75* N/A	0.08 0.21 0.43	0.04 0.20 0.63
Functional auditory performance (PEACH noise scale) 18 months 24 months 30 months	0.59 N/A N/A	0.57 0.23 N/A	0.42 0.32 0.16

*Correlation is significant at the 0.05 level

Phonological complexity of reported words

The total sum of the complexity parameters for each group showed that the parents of children with NH reported words with a slightly higher mean complexity at all ages compared to the parents of the children with HL: 18 months (NH = 2.66, HL = 2.48), 24 months (NH = 3.36, HL = 3.28) and 30 months (NH = 3.45, HL = 3.44), but this difference was not statistically significant. Individual mean complexity at 18, 24 and 30 months can be found in Table 4.

Discussion

The main purpose of this study was to investigate the expressive vocabulary development in children with moderate HL and compare their expressive vocabulary growth to a contemporary group of children with NH. Children with HL produced a similar number of words as the children with NH at 18 months, which is in line with previous studies (Moeller et al., 2007b). Although a positive finding, it may be expected to some extent due to the narrower range of number of words seen for all children around 18 months (Wordbank, 2020). However, the vocabulary growth from 18 to 24 months demonstrated a gap between the children with HL and NH, which widened even further from 24 to 30 months (the children with HL having fewer words). The increasing gap over time is in accordance with growth trajectories of expressive vocabulary in children with HL in previous research of children with various degrees of HL (Mayne et al., 2000; Vohr et al., 2011; Yoshinaga-Itano et al., 2017). This is of concern as the results from this study indicate that this may still be true even for children with moderate HL who have received HAs before the age of 6 months.

Another aim of the study was to examine the relationship between early consonant use and expressive vocabulary. The number of different true consonants at 18 months showed moderate correlations to the number of words produced at 24 and 30 months. These findings are in line with previous research (Moeller et al., 2007b; Stoel-Gammon, 2011). The observed discrepancy between number of different true consonants in relation to the number of produced words in individual children may in part be explained by that parents checked the words on the SECDI-II irrespective of how their child pronounced the word. The first words that children say are often used in a context which puts less need on the use of true consonants for parents to “hear and understand” what their child is saying. As the children with HL were observed to produce fewer (or no) true consonants at 10 and 18 months compared to the children with NH, and also produced fewer words at later ages, indicate a need to monitor early speech in this target group. It is also important to raise an awareness that although a child with HL vocalizes to the same extent as children with NH, the lack of true consonants may indicate a need for intervention (Moeller et al., 2007a; Stoel-Gammon, 1988). Furthermore, parental report on expressive vocabulary should be complemented with standardized assessments with an SLP.

With regards to the auditory variables, age at amplification did not show any significant association with number of words. However, the correlations increased with time indicating that the impact of early amplification may have a cumulative effect and may not be seen until later. Moreover, all children with HL were fitted with HAs before the age of 6 months, but six of the eight children were not full-time users until 10 months of age which could have affected the amount of their listening experience, and consequently delayed the children´s early use of speech. Hours of HA use and scores on auditory development were the auditory variables that showed significant correlations to number of produced words at later ages. Considering the variability seen in HA use the first three years highlights the need to motivate parents of infants with moderate HL to reach consistent HA use from an as early age as possible (Persson et al., 2020). The finding that all participants showed scores on auditory development within expected values, but few number of words produced was somewhat surprising as the construct of the LEAQ and SECDI-II have been found to be strongly correlated (Persson et al., 2019). However, this result could also be due to that the LEAQ is not as sensitive to milder degrees of HL (Bagatto et al., 2011; Persson et al., 2020).

Aided audibility did not have a significant impact on the number of words produced in this study. This is similar to the findings when correlating SII values to speech outcomes in children at 2 years of age (Ambrose et al., 2015). The main reason for this result is most likely due to the small cohort, albeit in line with previous studies demonstrating uncertainty in terms of what level of SII that may be “sufficient” for children at this young age degree of HL. However, as appropriate aided audibility is important for children with HL and the fact that it has been found to positively impact spoken language outcomes at later ages (Tomblin et al., 2014), future studies in larger groups of children are needed.

The relationship between functional auditory performance in noise and number of words produced was moderate but not significant. In addition to previous findings on the challenges of children with moderate HL to listen in noise (Persson et al., 2020), this raises a concern around these children’s ability to acquire new words as they spend many hours in facilities with fluctuating noise levels, and listening at a distance (Schafer et al., 2012). Young children who are at home may not face challenges with noise to the same extent as when they enter preschool, which puts larger demands on speech perception for all children. This calls for action to inform and support the professionals in preschools around the consequences of HL on spoken language acquisition.

The use of the WCM-SE to compare the complexity score of the parent-reported words was explorative as this measure is typically used to analyze samples of speech produced by the child. Based on previous findings that children with HL have specific challenges in producing certain details in speech (Ambrose et al., 2015), it could be hypothesized that the parents of the children with HL would report their children to use words with lower complexity scores. As the words produced were based on parental report it is not possible to give any statements around the abilities of expressive phonology of the children, but the similar complexity between the groups indicates that the children with HL at least had acquired words with similar complexity as the children with NH. As stated in previous sections, it could be that parents of the children with HL understood their children´s word productions despite their lower number of different true consonants, due to context. Therefore, it is recommended that the phonological complexity of the children should be investigated based on their true word productions. A standardized test of early expressive phonological skills in children between the ages of 18–30 months is under development (Profiles of Early Expressive Phonological Skills, PEEPS) (Stoel-Gammon & Williams, 2013) as well as a Swedish version (PEEPS-SE) (Marklund et al., 2018b). Such a test could possibly fill the gap in the assessment of early expressive phonology in young children.

Strengths and limitations

The main limitation of this study was the limited number of participants and the authors acknowledge the constraints in statistical analyses and recognize the findings to be taken with caution, rather than applying the results to the population at large. Therefore, larger samples are encouraged in future studies. Yet, the limited number of participants included all eligible children with the set inclusionary criteria during the recruitment period, except one. The study also controlled for several factors known to affect vocabulary outcomes in previous largescale studies gives strength to the results.

This study utilized well-established as well as newer measures. The Swedish CDI has been extensively used in previous research; however, the use of the Swedish Word Complexity Measure was explorative. There is a need of further trials with larger groups to evaluate its usefulness. Although the SECDI is an acknowledged tool, the data on the number of produced words were only collected through parent-report which could have benefit from complements of standardized tests, if there would have been any ones available. This highlights the need for instruments of this kind for this particular age group.

The experiences from the measures used in this study has indicated to be clinically relevant as they are feasible and available in Swedish, albeit not currently used in the monitoring of Swedish children with moderate HL at a national level.

Conclusion and future directions

The findings from this study are in line with previous research demonstrating that children with moderate HL are at-risk for future delays in their expressive vocabulary development despite early fitting of HAs. To support the development of expressive vocabulary, early intervention services of children with HL should promote full-time use of HAs, assessments of early consonant production, and include strategies to alleviate the challenges of listening in noise. As the gap in expressive vocabulary increased with age, monitoring development at later ages are also encouraged. Future studies need to include larger groups of children and standardized assessments of expressive vocabulary as well as video (audio) recordings of words.

Acknowledgements

The authors would like to thank all the participating children and their families for their contributions to this longitudinal study. Special thanks to all colleagues involved in data collection and assessments. This work would not have been possible without the financial support from Wibelfonden, Tysta skolan, Majblomman, and Region Stockholm (Medical Unit Speech and Language Pathology).

Disclosure statement

The authors have no conflict of interest to declare.

Additional information

Funding

This study was supported by funding from Wibelfonden, Tysta skolan, Majblomman and Region Stockholm.