Assessing the Brighton Collaboration Case Definition of Sensorineural Hearing Loss in Ghana

ABSTRACT.

The Brighton Collaboration has developed a case definition to assess sensorineural hearing loss (SNHL) as an adverse event of special interest for Lassa fever vaccines. We applied the level of diagnostic certainty (LOC) criteria to 300 SNHL patients aged 18 to 59 years at a tertiary hospital in Ghana from January 2017 through June 2020 and evaluated the applicability of this definition. Most SNHL cases were assessable (85.0%) and assigned level 1 LOC (84.3%); missing information on otoscopy (86.7%) was the main reason for being unable to classify cases. Consistency of LOC classification between assessors was 99.3%. Cases with electronic medical records (EMRs) were less assessable than those with paper records (30.9% versus 93.8%). These findings indicate that the SNHL definition would be applicable to retrospectively ascertain and classify cases in resource-limited settings. Developing an EMR template to document otoscopy results may improve the feasibility at this hospital to ascertain SNHL.

INTRODUCTION

Lassa virus is a primary cause of hemorrhagic fever worldwide and endemic over most of West Africa. In West Africa, annually 100,000 to 300,000 cases of Lassa fever (LF) have been estimated, resulting in approximately 5,000 deaths.¹ A third of LF survivors develop unilateral or bilateral sensorineural hearing loss (SNHL), from which only some patients fully recover.^2,3 The SNHL cases are identified at 10 to 15 days after the onset of symptoms or at the convalescent stage, suggesting that it could be due to direct viral damage, immune-mediated process, or a combination of both.³ Permanent SNHL carries a huge cost burden economically and socially. The disability and stigma of deafness lead to higher absenteeism and lack of employment opportunities; social isolation and impaired communication; and psychological distress.³ In low- and middle-income countries, a patient’s quality of life is further exacerbated because treatment such as hearing aids or cochlear implants is not usually administered.

Multiple LF vaccine candidates are in the development pipeline.⁴ To identify potential safety issues that may arise in LF vaccine trials, the Safety Platform for Emergency vACcines (SPEAC) project has created a list of potential adverse events of special interest in 2019, which includes SNHL.⁵ A Brighton Collaboration case definition for SNHL was then developed and made available in February 2020.⁶ The Brighton Collaboration case definitions aim to enhance comparability of immunization safety data and are designed to accommodate varying levels of diagnostic certainty (LOC) based on available information. This is achieved by a case definition format allowing for three LOC from highly specific (level 1) to a stepwise increase of sensitivity (level 3), plus two categories for events with insufficient information (level 4) or are rejected (level 5).⁷ Case reports of SNHL after vaccination for influenza, measles/mumps, and other diseases have led to concerns that immunization may rarely cause hearing loss.^8–11 Beyond case reports, however, a large-scale analysis of sudden SNHL after more than 20 million influenza and other vaccines did not detect any association in the 1 to 28 days postvaccination.¹² Recently, a cohort study of 2.6 million patients from Israel suggested a slightly higher risk of sudden SNHL within 21 days after receipt of BNT162b2 mRNA COVID-19 vaccine.¹³

There is a renewed desire to ensure that LF vaccine trials take place in affected countries. In Ghana, serologic evidence of Lassa virus infection and indigenous human cases have been reported^14,15; the country also has investigational sites identified as ready to host LF vaccine trials immediately or with little support.¹⁶ Given the potential use of the SNHL case definition in LF and other emerging vaccine trials, it is important to evaluate its field performance for revision or adaptation, particularly in resource-limited settings. This study retrospectively assessed the applicability of this definition to data collected in routine clinical care settings at a tertiary hospital in Ghana.

MATERIALS AND METHODS

Study setting.

The Komfo Anokye Teaching Hospital (KATH) is the second largest teaching and specialized hospital in Ghana.¹⁷ Located in Kumasi, the Regional Capital of Ashanti Region with a projected population of 4.7 million, this 1,200-bed hospital has more than 4,000 permanent staff and 15 directorates and takes direct referrals from 12 of 16 administrative regions in Ghana. Being on an international road connecting the northern part of Ghana to the capital city Accra, KATH also receives patients from neighboring countries such as Burkina Faso and Côte d’Ivoire.

Data collection.

The study subjects were patients aged 18 to 59 years who attended the KATH Eye, Ear, Nose and Throat (EENT) department between January 1, 2017 and June 30, 2020. Patients of SNHL were retrospectively identified through electronic (Lightwave Health Information Management System¹⁸ [LHIMS]) and paper medical records.

LHIMS is a web-based electronic medical records (EMRs) system implemented by KATH starting January 2019 and rolled out in all departments and directorates; the EENT department was actively using the LHIMS starting in January 2020. The LHIMS database has 20 interconnected subsections, which include vital signs, nurse notes, history, principal complaint, (general) on direct questioning, (general) examination, drawing pad, diagnosis-related group or International Classification of Diseases helper, treatment plan, inventory debit, prescriptions, recommendations, laboratory parameters, doctor notes and remarks, admission recommendations, audiological evaluation, attached documents, images and videos, and attach signature. The LHIMS interface has other components for searching, data extraction, preview, and patient profile and demographic characteristics.

The principal investigator (I. B.) conducted training for five junior officers at the EENT department to assist in case identification and data abstraction. Using the keywords “hearing loss”, “sensorineural hearing loss”, or “SNHL”, two officers manually searched consulting room registers (January 1, 2017–December 31, 2019), and the head of biostatistics unit searched LHIMS through the 20 subsections (January 1–June 30, 2020) to generate a list of potential patients. The list included, for each subject, study identification number, hospital registration number (unique personal identification number), age, gender, region, principal diagnosis, and eligibility/reasons for exclusion. Another two officers retrieved patient medical files from the records room for review.

A data collection form was developed and used to collect information on the key criteria of LOC ascertainment⁶: 1) patency of the ear canal plus visible, intact, and mobile tympanic membrane with no evidence of middle ear disease at otoscopy; 2) normal (type A) tympanogram; 3) audiometry consistent with SNHL; 4) tuning fork exam consistent with SNHL; 5) auditory brainstem response (ABR) consistent with SNHL; 6) abnormal otoacoustic emissions (OAE); 7) abnormal behavioral or neurodevelopmental testing questionnaires; and 8) abnormal remote screening. A total of 300 subjects were selected from potential patients, by date from the most recent patient backward. The principal investigator (I. B.) and an officer abstracted data from medical records of each of the 300 patients, entered the data into a password-protected spreadsheet independently, and crosschecked data validity and accuracy.

Per SNHL case definition (Table 1),⁶ a Brighton Collaboration LOC was assigned to each abstracted case by each abstractor; where there were discrepancies, a third person not involved in data collection (W.-T. H.) reviewed the case and verified the LOC.

Table 1.

The Brighton Collaboration criteria for level of diagnostic certainty ascertainment of an SNHL case⁶

Level 1

A physical examination excluding conductive hearing loss

AND

Audiometry consistent with SNHL

Level 2

A physical examination excluding conductive hearing loss

AND

Auditory brainstem response test consistent with SNHL

OR

Tuning fork exam consistent with SNHL

Level 3

A physical examination excluding conductive hearing loss

AND

Otoacoustic emissions test consistent with hearing loss

OR

Behavioral or neurodevelopmental testing questionnaire concerning for hearing loss

OR

Remote screening using telehealth technology concerning for hearing loss

Level 4

Insufficient information to diagnose SNHL

Level 5

Not a case of SNHL

SNHL = sensorineural hearing loss.

^*Case definition of SNHL: SNHL is hearing loss of ≥ 30 dB in three sequential frequencies in the standard pure tone audiogram.

^†The physical examination must include otoscopy, which is necessary to evaluate the patency of the ear canal and the condition and movement of the tympanic membrane. A tympanogram, if available, could be included as part of the physical examination to exclude conductive hearing loss, but it is not considered essential.

Statistical analysis.

Using the deidentified datasets, sociodemographic and the following parameters were assessed descriptively: 1) completeness to obtain information on each criterion of the case definition, 2) ability to assign LOC up to level 3 and what LOC could be ascertained, 3) ability to achieve the same LOC for the same case by different assessors, and 4) reasons for being unable to classify cases. Differences between EMR and paper-registered cases were compared using χ² or Fisher’s exact tests.

RESULTS

Overall, 860 potential cases were identified; records from the most recent 300 cases (42 electronic and 258 paper) were abstracted. Patients with EMRs were more likely to be male (P = 0.132), married (P < 0.001), have no formal schooling (P < 0.001), and work as vocational workers (P = 0.004), compared with those with paper records (Table 2).

Table 2.

Sociodemographic characteristics of patients with SNHL by data source

Characteristic	Paper (N = 258)	Electronic (N = 42)
Male gender, n (%)	137 (53.1)	28 (66.7)
Age, y, median (interquartile range)	43 (22.0–55.0)	43 (34.0–50.8)
Marital status, n (%)
Never married	95 (36.8)	9 (21.4)
Married	101 (39.1)	25 (59.5)
Divorced or widowed	1 (0.4)	3 (7.1)
Unknown	61 (23.6)	5 (11.9)
Highest education attained, n (%)
No formal schooling	0 (0.0)	6 (14.3)
Primary or junior high school	17 (6.6)	8 (19.0)
Senior high school	55 (21.3)	13 (31.0)
College and above	43 (16.7)	4 (9.5)
Unknown	143 (55.4)	11 (26.2)
Occupation, n (%)
Civil servants	34 (13.2)	7 (16.7)
Vocational workers	45 (17.4)	15 (35.7)
Students	63 (24.4)	2 (4.8)
Farmers	26 (10.1)	2 (4.8)
Unemployed	14 (5.4)	0 (0.0)
Unknown	76 (29.5)	16 (38.1)

SNHL = sensorineural hearing loss.

SNHL was mainly assigned level 1 LOC (Table 3). The consistency of LOC classification between the two abstractors was 99.3% (298/300; two level 2 cases assigned as level 4). Forty-five cases were nonassessable; 39 (86.7%) were due to missing information on otoscopy. Cases with EMRs were less assessable than those with paper records (30.9% versus 93.8%, P < 0.001).

Table 3.

Level of diagnostic certainty and documented evidence of criteria for ascertainment by data source

	Paper (N = 258)	Electronic (N = 42)
Level of diagnostic certainty, n (%)
1	240 (93.0)	13 (30.9)
2	2 (0.8)	0 (0.0)
3	0 (0.0)	0 (0.0)
4	16 (6.2)	29 (69.1)
5	0 (0.0)	0 (0.0)
Criterion, n (%)
Otoscopy	242 (93.7)	16 (38.1)
Tympanometry	241 (93.4)	16 (38.1)
Audiometry	240 (93.0)	13 (30.9)
Tuning fork exam	2 (0.7)	0 (0.0)
Auditory brainstem response test	0 (0.0)	0 (0.0)
Otoacoustic emissions test	0 (0.0)	0 (0.0)
Behavioral or neurodevelopmental assessment	0 (0.0)	0 (0.0)
Remote screening	0 (0.0)	0 (0.0)

Completeness on the recorded evidence of otoscopy (P < 0.001), tympanometry (P < 0.001), and audiometry (P < 0.001) was lower in EMRs than in paper records (Table 3). Tuning fork exam, ABR, OAE, behavior/neurodevelopmental assessment, and remote screening were rarely documented. Interviews of EENT physicians confirmed that neither ABR nor OAE were available at KATH, and behavioral questionnaires or remote screening via telehealth technologies were not routinely used to detect hearing loss.

DISCUSSION

Findings from this study indicate that more than 80% of the SNHL cases at KATH, a tertiary hospital in Ghana, could be assigned a level 1 LOC retrospectively. When a case was not assessable, missing data on otoscopy was the main reason. The very high degree of consistency between assessors (> 99%) and the fact that 93% of the paper records abstracted met level 1 LOC indicate that this definition would be applicable to ascertain and classify cases in resource-limited settings. However, the fact that only 31% of cases meeting level 1 LOC through EMRs would greatly influence the feasibility at this site to conduct surveillance for SNHL in the context of an LF vaccine trial or post-introduction.

Otoscopic visualization of the ear canal and tympanic membrane to exclude conductive hearing loss is considered essential to ascertain SNHL.⁶ A substantial proportion (87%) of nonassessable cases lacked this information in their records, and those with EMRs had lower completeness of documentation. This could be due to a recent change of patient management procedures (e.g., no such examination being performed) or more likely, because of poor documentation of key LOC criteria in EMRs. The EENT department shifted to the LHIMS platform in January 2020 and since then, medical records of each patient were maintained in digital format. The physicians were possibly not familiar with the EMR system and hence did not enter or record those notes. Without parallel paper systems, however, we were unable to assess whether more data might be incorporated into EMRs should paper records abstraction be done. Use of a checklist template to include otoscopy for the LHIMS examination subsection may improve the overall documentation quality.

This study has several limitations, including the restriction to only adult patients aged 18 to 59 years. Although this population may be first enrolled in an LF vaccine trial, it underrepresents the elderly in which occurrence of SNHL is higher¹⁹ and may not be generalizable to pediatric patients. Next, our findings are limited to a tertiary hospital in Ghana; whether they could be generalized to other hospitals with lower referral levels in the country or elsewhere in Africa remains unclear. Lastly, SNHL cases were identified retrospectively; we did not apply and evaluate the Brighton Collaboration criteria for LOC ascertainment in prospective studies.

Although performing an audiometry is feasible to ascertain a level 1 LOC at KATH, the required ABR or OAE tests to assign patients to levels 2 or 3 are not available at this hospital; instead, they are offered at private hospitals. Rarely, tuning fork exams were documented in medical records. Training the testers at district, municipal, and regional hospitals to use tuning forks in routine evaluation of hearing impairment could be a complementary way to better identify SNHL in these low-resource settings.