Article Text
Abstract
Background Systematic appraisal of universal newborn hearing screening (UNHS) programme performance and the prevalence of permanent childhood hearing impairment (PCHI) detected is lacking, including for those admitted to Neonatal Intensive Care Units (NICU). We carried out a systematic review and meta-analysis of studies reporting PCHI prevalence (defined as bilateral loss ≥26 dB HL) detected through UNHS (defined as universal screening using otoacoustic emissions and/or auditory brainstem response testing by age 6 months) in very-highly developed countries (PROSPERO:CRD42016051267). We estimated screening test performance and PCHI prevalence in those who were or were not admitted to NICU.
Methods Eligible studies, reporting UNHS-detected PCHI prevalence in very highly-developed countries (no restrictions by language or date), were identified from six electronic databases (January 2017) along with references of cited papers and unpublished literature (November 2017). Papers reporting on at-risk populations only, with no English abstract (unless unpublished), or of ineligible study type were excluded. Two reviewers independently extracted data and assessed quality of included papers using criteria adapted from the Newcastle-Ottawa, STARD and QUADAS-2 tools, with differences resolved by consensus. Pooled prevalence was estimated from random-effects models using Freeman-Tukey double arcsine transformation. Negative predictive value (NPV), sensitivity and specificity were calculated only for studies with follow-up to ascertain false negatives, whilst positive predictive value (PPV) calculation was not restricted by follow-up. Confidence intervals (95% CI) were estimated using Wilson (Score) methods (Stata: Release 15; StataCorp LP).
Results 41 eligible reports on 32 study populations (1,799,863 infants) were identified from 6195 non-duplicate references. Pooled UNHS-detected PCHI prevalence was 1.08 (95% CI 0.90 to 1.28) per 1000 screened (I2 =89.2%). Prevalence was 6.9 times (95% CI 3.8 to 12.5) higher among those admitted to NICU (3 studies). Smaller studies were significantly associated with larger prevalences (Egger’s test: p=0.017). PPV ranged from 1.5%–83.5% (25 studies), NPV 100% (7 studies), sensitivities 88.9%–100% (8 studies) and specificities 92.3%–99.9% (7 studies). Quantitative pooling of screening programme performance was not possible due to methodological differences.
Conclusion In very highly-developed countries, around 1 per 1000 screened infants will require PCHI investigation and management. Prevalence is almost 7 times higher in infants admitted to NICU. Strengths of our study include the systematic search strategy and robust statistical methods. Our findings are limited to very highly-developed countries. Estimates were restricted by lack of high-quality reporting on attrition and surveillance. Improved reporting of surveillance and attrition should be encouraged to enable evaluation of screening programme performance.
ESRC-funded PhD ES/J500185/1.