Predictive model for progression of hearing loss: meta‐analysis of multi‐state outcome

Abstract

To develop a multi-state Markov model to predict multi-state progression of age-related hearing loss (ARHL). A systematic searching of literature from Medline (1966-2005) was performed. The disease process for hearing loss was modelled as a four-state continuous-time Markov process. We estimated the progression rates for each study separately, then calculated weighted averages over all studies across age groups, weighting for each study by the inverse of variance. The pooled estimates were obtained and transition probabilities between states were illustrated. The progression of hearing loss increased with the ascending frequencies across all age groups. Men had significantly faster progression rates in all frequencies and age groups except for the age group of 90 years or older. In comparison between ears, the progression of hearing would be slightly faster in left ears initially in early elder life and did not show any difference in further ageing and later hearing declines. With the pooled estimates of progression rates, the probabilities of hearing deterioration could be obtained. The multi-state model can quantify the nature course of hearing progression in ARHL. Predictions of hearing status can be simulated either at population or individual level with this model.