This letter comments on the letter by Abiola Aanuoluwa. To the Editor: In our recently published study,1 we used administrative claims data to examine the association between hearing aids (HAs) and dementia among older adults with hearing loss (HL). In 2017, a Lancet Commission report highlighted that HL carried a greater risk of dementia than other potential risk factors, such as hypertension and obesity, and could be expected to have a high population‐attributable risk due to the increasing prevalence of HL.2 The study further calculated that more than one‐third of dementia cases may be prevented by addressing risk factors across the life course. HL is a potentially modifiable risk factor for Alzheimer's disease and related dementias.2 Although randomized clinical trials may still be considered the gold standard in clinical research, with the availability of large data sets, observational studies are increasingly gaining momentum. Observational studies are inherently less expensive, include a broader patient population, are more efficient, and can examine longitudinal data over a relatively long period of time.3 With regard to potential biases, there is little evidence to support the superiority of randomized clinical trials over observational studies.3 In our recent study, our descriptive results revealed that in aggregate, people with HAs had lower rates of cardiovascular conditions, hypertension, hypercholesterolemia, obesity, and diabetes at the time of HL diagnosis. Abiola et al raised a valid concern regarding a potential selection bias in our study.4 All studies, particularly observational studies, are prone to selection bias when comparing different population groups. To address this, we adjusted our analytical models for age, sex, race/ethnicity, and chronic conditions that are more prevalent among individuals with HL and conducted several sensitivity analyses including propensity score matching between those with and without HL. None of our sensitivity analyses qualitatively changed the results of our original analyses. Abiola et al also raised a question regarding timing of HA use.4 Because we were interested in the association between HAs and time to diagnosis of certain conditions, we included those who acquired HAs right after HL diagnosis (about 60% of users of HAs). This made our analysis much cleaner because we were able to follow each patient for at least 3 years after the diagnosis of HL and first use of HAs. It is plausible that by excluding those who had fewer than 3 years of follow‐up, we introduced some bias into our analysis. Not having information on why an individual is no longer covered by a health plan (eg, death, switching health plans, switching jobs, etc), we examined the results for those with at least 3 years of follow‐up time. The main shortcoming of our study was unavailability of information on several salient risk factors in our data set. For example, we did not have information on duration and severity of HL, frequency of HA use, if any, socioeconomic status, and lifestyle choices of our patient population. Future research using other data sets should try to address the limitations of our study and therefore confirm or dispute our findings ...
Language: en
LA - en SN - 0002-8614 UR - http://dx.doi.org/10.1111/jgs.16366 ID - ref1 ER -