Sample
The present study combines baseline data from the School Hearing Investigation in Nord-Trøndelag (SHINT) with follow-up data from the Nord-Trøndelag Hearing Loss Study (NTHLS) and the second wave of the Nord-Trøndelag Health Study (HUNT 2) in Norway. All NTHLS participants also participated in the HUNT 2. Altogether, data from three questionnaires are used in this study: the questionnaire from the NTHLS (NTHLS-Q1), and both questionnaires from HUNT 2 (HUNT2-Q1 and HUNT2-Q2). The HUNT2-Q2 was handed out at the examination cite and returned by mail later, the response rate for the sample used in the present study was 76.3%.
The SHINT was an audiometric screening of all schoolchildren attending regular schools in the county of Nord-Trøndelag aged 7, 10 and 13 years from 1954 to 1986. The late H.F. Fabritius who was a Norwegian Ear, Nose and Throat (ENT) specialist, led the investigation. Great efforts were made to ensure participation, and Dr. Fabritius himself even rowed a boat to reach a small island off the Norwegian coast in order to include the people who lived there. Unfortunately, the exact number of participants is unknown, since records were only made for children with hearing loss, and not for children with normal hearing. We do know, however, that 78,524 children were born in Nord-Trøndelag between 1941 and 1977, which may serve as a proxy.
The initial screening included as good as every single pupil in the entire county and took place in a quiet location at the respective school. A trained hearing assistant or a nurse performed the hearing examination. Air-conduction thresholds were obtained by means of pure tone audiometry at 0.25, 0.5, 1, 2, 4 and 8 kHz utilizing Amplivox audiometers (type 70, and later models 51 and 81). Pupils were registered with hearing loss if 1) thresholds of 20 dB or greater at three or more frequencies in the same ear were detected, or if 2) a threshold of 30 dB or more at one or more frequencies were detected.
A total of 10,269 children were classified with hearing loss at the screening. All of these children were then invited to a full examination by an ENT specialist at one of the 14 out-patient clinics in Nord-Trøndelag. Questionnaire data regarding the children’s ear problems were collected from the parents. The ENT specialist performed a new pure tone audiometry with both air- and bone-conduction thresholds as well as a complete medical examination including family and medical history, recording findings and diagnoses. Children underwent one or more ENT examinations depending on the diagnoses in order to ensure correct classification. Dr. Fabritius defined SNHL as hearing loss in which the air-conduction thresholds followed those of the bone-conduction, although he did not include a maximum accepted air-bone gap in this definition. In this study, we rely on the diagnosis made by Dr. Fabritius and the other ENT specialists. The attendance rate was 97% between 1954 and 1962 and it is likely that the high participation rate persisted [
22].
Out of the 10,269 children who tested positively for hearing loss at the screening, 1489 were diagnosed with Sensorineural Hearing Loss according to Dr. Fabritius’ definition. However, only 3066 out of the 10,269 children from the screening in the SHINT also participated in the NTHLS as adults, and just 462 out of the original 1489 SNHL cases from the SHINT also participated in the NTHLS. There were several reasons for this attrition, like for example loss of identification number or not being old enough to be invited to the NTHLS, or possibly moving away from the county (for more details, see [
23]).
For the purpose of the present study, we wanted to distinguish between profound-severe, moderate, and mild hearing loss, respectively. We estimated the average hearing threshold of 0.5, 1, 2 and 4 kHz in both ears from the last audiogram (from the ENT examination, not from the screening), which for most participants was at age 13. This means that the hearing loss might have emerged at different points in time for the participants, somewhere between birth and 13 years of age. We defined moderate-severe hearing loss as 41 dB or more (ranging to 100, which means that this group also includes profound hearing loss), mild hearing loss as 26–40 dB, and slight hearing loss as 16–25 dB, resulting in 67 cases in each of the two former groups and 223 cases in the latter. This reduced the case group from 462 to 357.
Finally, since the present study has mental health variables as outcome, we wanted to exclude cases that might be struggling with mental health problems at baseline. We cross-checked the case group with data on the following conditions registered at baseline: “Retarded”, “Cerebral paralysis”, “Mental health issues”, “Mental distress/ depressed”, “Down’s syndrome”, and “Is receiving psychological treatment”. Three individuals were registered as “Retarded”, one individual was registered with “Cerebral paralysis”, and one individual with “Downs syndrome”. These cases were excluded from further analysis, resulting in a total case group of 357 of which 220 are in the case group with slight hearing loss whereas the other two groups include 66 individuals each.
The NTHLS was a part of the second wave of the Nord-Trøndelag Health Study (HUNT 2) carried out in 1995–97. In HUNT 2, the entire adult population in Nord-Trøndelag County was invited to participate, whereas in NTHLS, the adult population in 17 of the 23 municipalities was invited to participate. Data on mental health were available for 51,574 people (62.8%) and the age span was 20 to 101 years. Participants in the NTHLS who had not been diagnosed with a hearing loss at SHINT were used as the reference group, following the assumption that all of these people grew up in Nord-Trøndelag and therefore participated in the SHINT. Since the SHINT lasted from 1954 to 1986 and the NTHLS from 1995 to 1997, the oldest participants to attend both SHINT and NTHLS would have been 13 years in 1954, thus 56 years old in 1997. Therefore, we only selected people 56 years and younger from the NTHLS for the reference group. This resulted in a sample of 32,456 individuals.
To sum up, the sample consisted of 66 individuals with moderate-severe hearing loss, 66 with mild hearing loss, 220 with slight hearing loss, and 32,104 with normal hearing (reference group).
Measures
Childhood sensorineural hearing loss (predictor)
For the purpose of the present study, only children diagnosed with SNHL were selected. In the present study, we estimated the average hearing threshold of 0.5, 1, 2 and 4 kHz in both ears from the last audiogram. We defined moderate-severe hearing loss as 41 dB or more, mild hearing loss as 26–40 dB, and slight hearing loss as 16–25 dB.
Mental health (outcome)
Ten of the 25 items from the Symptom Checklist-25 [
24], here called SCL-10, were included in the NTHLS-Q1 and were used to measure mental health. Four questions tap anxiety and six questions tap depression. The distribution of the SCL-10 scores was skewed and the scores were therefore log transformed. A high score on this index reflects poor mental health. Cronbach’s alpha was .86. Using another available data material [
25], we estimated the correlation between the SCL-25 global score (anxiety and depression jointly) and the short-form global score to .97.
Subjective well-being (outcome)
The index consists of three items from HUNT2-Q1, phrased as follows: When you think about your life at the moment, would you say that you are by and large satisfied with life, or are you mostly dissatisfied? (seven response categories ranging from “very happy” to “very unhappy”); In the course of the last 2 weeks, have you been feeling safe and calm?; In the course of the last 2 weeks, have you been feeling happy and optimistic? (four response categories ranging from “no” to “a lot” for both items). The first question was recoded so that a high score on this variable reflects a high level of subjective well-being. Because of the different number of response categories, the items were standardized before they were added into a sum score indicator. Cronbach’s alpha was .84.
Self-esteem (outcome)
Four items from The Rosenberg Self-Esteem Scale [
26] were included in HUNT2-Q2. The questions are phrased as follows: I take a positive attitude toward myself; I feel that I am a person of worth, at least on an equal plane with others; I feel I do not have much to be proud of; I certainly feel useless at times. The two last items were recoded before the items were added as a sum score so that a high score on this variable reflects a low level of self-esteem. Cronbach’s alpha was estimated to.74 in the present data set. The four-item short form scale has been shown to correlate .95 with the original instrument [
21].
The scores for all of the three outcome variables were standardized before entered in the analyses.
Control variables
Control variables included in this study are age, education, mother’s education and father’s education, respectively. The data were provided by Statistics Norway with the following ten categories: 1) No education/preschool, 2) primary school 1st to 7th grade, 3) middle school 8th to 10th grade, 4) high school 11th to 12th grade, 5) high school diploma 13th grade, 6) high school extended, 7) college or university, lower level, 14th to 17th grade, 8) college or university, higher level, 18th to 19th grade, 9) PhD level, 20th grade or more, 10) education not reported. We recoded categories 1, 2, 3 and 10 into “Primary school”, category 4 into “middle school”, categories 5 and 6 into “high school”, category 7 into “college/university, less than 4 years”, and categories 8 and 9 into “college/university, 4 years or more”.
Treatment of missing values
As mentioned earlier, records for normal hearing were not registered in the SHINT, which means that for many frequencies, values below 20 dB were missing. The missing value for each frequency was therefore replaced by the frequency specific mean values of the scores below 20 dB in the original sample (N = 10,269).
For the outcome variables, we used SPSS Missing Value Analysis (MVA), expectation maximization (EM) for imputation of missing data where the respondent had valid data on at least half of the items. The ten SCL items were used to predict each other. This reduced missing values from 7.9 to 0.3%. For SWB, the items were used to predict each other in those cases where the respondent had valid data on two of the three questions, reducing missing values from 8.0 to 4.9%. The Self-esteem items were included in HUNT2-Q2, and, as mentioned above, this questionnaire was returned by somewhat fewer respondents (76.3%) than the other two questionnaires. For this reason, there was a larger percentage of missing values on the Self-esteem variable compared to the SCL-10 and SWB variables. Missing data were replaced for respondents who had valid data on at least two of the four items. The four Self-esteem items were used to predict each other, reducing missing values from 19.6 to 17.8%.
A total of 319 participants (0.9%) as well as 4697 fathers (14.5%) and 3402 mothers (14.5%) did not report level of education and missing values were replaced by the sample mean.
Design and statistical analyses
This study applies a longitudinal design, investigating the association between childhood sensorineural hearing loss at baseline and adult mental health up to 43 years later. In order to study this association separately in men and women, we split the dataset into to new datasets; one including men only, and one with women only. Three ANOVA analyses (IBM SPSS 24, General Linear Models, Unianova) were conducted consecutively in each data set (the total sample, the male sample and the female sample, respectively) with Childhood Sensorineural Hearing Loss (CSNHL) as the predictor and SCL-10, Subjective Well-Being, and Self-Esteem as the respective outcomes. Since the dependent variables were standardized before entered in the analyses, the unstandardized regression coefficients (b) show adjusted group mean differences scaled in fractions of a standard deviation. This makes it easier to interpret the results. The first model tests the association between childhood sensorineural hearing loss and adult mental health, whereas the second model tests the same association controlled for age, education, and mother’s and father’s education, respectively. In the analyses with the total sample we also included sex as a control variable (but not in the analyses with the male sample or female sample).