To our knowledge, this is the first study that demonstrates the existence of a RS in CI users. We found a positive RS for all areas, reaching statistical significance in the total score, in the general domains psychological and social and in the subdomains SE and AL.

To detect RS, we used the then-test method, which is often criticized for being susceptible to recall bias [29, 40] and may contain noise [41]. The deactivation of the CI helps participants to recall their preoperative state and thus minimized bias and noise. However, this allows the possibility that CI users do not in fact think back, but evaluate their temporary present state under the condition with a deactivated CI. In the cases where preoperatively any residual hearing was present, this may be preserved despite the implantation, however, is lost in the majority of cases [42]. Nevertheless, we found that the pre-t1 (then-test) results were equal or better than the t0 preoperative results. This leads us to the conclusion that the preoperative state actually dominated their then-test responses.

A RS in HRQoL has been investigated in numerous categories of diseases, from mild to life-threatening, including chronic and terminal conditions [26]. Cochlear implantation requires surgery and rehabilitation, after which CI recipients oftentimes can go back to their previous lives before hearing loss.

Korfage et al. [27, 43] found evidence for a RS in men with prostate cancer. The studies indicated that the RS was primarily induced by the cancer diagnosis per se and was larger in size than that induced by the treatment. However, the ES were small to negligible. In people with functional deafness, the diagnosis does not provide them with fundamentally new information due to the participants’ apparent knowledge that their hearing had reduced. Thus, a RS induced by the diagnosis hearing loss cannot occur. In comparison to the prostate cancer studies, we found that the ES of the treatment RS (B) was larger within our study. The direction of the RS (then minus pre) in the prostate cancer patients study was negative, perhaps due to a deterioration in health, however the authors stated it as interpretable for prostate cancer patients [27].

Bernhard et al. [44] evaluated the then-test method by investigating ten QoL indicators via visual analog scales (VAS) in two consecutive treatments in colon cancer patients, first a radical resection surgery and second a post-operative adjuvant chemotherapy composed of three randomized treatment arms, one of them observation. The resection surgery generally resulted in a worsening (i.e., negative observed change) and negative RS in the physical indicators “physical well-being”, “tiredness”, and “functional performance”, while causing improvement (i.e., positive observed change) and no significant RS in the psychological indicators “mood”, “perceived adjustment”, and “anxiety”. The adjuvant treatment arms showed no (6 indicators), positive (3 indicators) or inconsistent between treatments (2 indicators) observed change, generally smaller than the changes around resection surgery. As a result of the negative RS, the then-test change was positive in 9 out of the 10 indicators. However, selectivity between treatment study arms did not improve.

In contrast to the physical indicators regarding surgery treatment in Bernhard et al. [44], in our study the observed changes were significantly positive, and the RS effect sizes were positive, though small. This may be due to the physical indicators in Bernhard et al. [44] addressing physical health state shortly after surgery, whilst our study addressed functional hearing performance after the consolidation phase in hearing rehabilitation. However, both studies showed a RS that followed the direction of the observed change. Psychological indicators yielded improvement in the observed change in both studies, however the RS was significantly positive with medium ES in our study and not significant in the study of Bernhard et al. [44].

Thus, when considering the observations of Bernhard et al. [44], utilizing the then-test method for selectivity between treatments appears questionable. The CI is currently the only existing treatment for rehabilitation of functional deafness, so the question doesn't arise, but that may change in the future.

People with chronic illness type I diabetics with unsuccessful pancreas and kidney transplant procedures, retrospectively overestimate their pre-transplant QoL [45]. Adang et al. [32] demonstrated that type I diabetics with successful transplant procedures underestimate their pre-transplant QoL. These observations suggest that the RS could be outcome dependent. Similar to our results, Adang et al. [32] found significant positive observed change, sustained over an extended period of time after the surgical intervention. The RS was negative though, analogous to other life-threatening conditions, such as cancer [27, 44].

Joore et al. [33] demonstrated that a RS in QoL was present in hearing impaired adults after hearing aid fitting. The RS was primarily seen in the hearing related QoL dimensions and not the generic dimensions that were used as control. Although it could be assumed that hearing aids and CIs are comparable regarding hearing improvement, we in contrast primarily observed a RS in the psychological and social domains. Moreover, the RS observed by Joore et al. [33] were negative with hearing aids, whereas we found positive RS after CI provision.

Joore et al. [33] mention the reluctance of many older persons to accept the necessity of a hearing aid. Their pre-treatment responses could therefore be more positive (since they are not realistic) than their post-treatment responses (which should better match reality), thus resulting in a negative RS. A substantial difference between the groups is the severity of the impairment: while many people can cope with the hearing handicap caused by mild to moderate hearing loss (which can still be improved by conventional hearing aids), unaided functional deafness is more difficult to compensate and/or ignore. At the time of the t0 pre-test, the final decision for CI surgery had already been made and a still unrealistically positive assessment of one’s hearing status is unlikely. In comparison, the prescription and prospective use of a hearing aid for an extended time may appear much less severe and final and thus still leave room for handicap denial.

The mean age of participants in the Joore et al. study [33] was 67 years (± 12), while in our cohort the mean age was 63.4 years (± 15.8), rendering the age too similar to explain the diverging differences in the RS. A possible difference could be that the participants within our study were advised to temporarily deactivate their CI to help them recall the preoperative state.

In our study, the RS was positive, likely because the participants raised their standards as a result of hearing improvement and possibly also hope for further improvement with extended use.

According to the meta-analysis conducted by Schwartz et al., the size and the direction of the RS vary considerably in the literature [26]. In this meta-analysis the authors discuss, that in the case two studies investigating similar circumstances reveal different RS directions, then the sign of the RS is relevant and should lead to questioning the validity of the findings [26]. In our study we found that the largest RS had occurred within the social domains and the largest ES in the social and psychological domains. This is in accordance with the hypothesis of Schwartz & Sprangers [24] that a RS is more likely to occur within subjective rather than objective domains.

Regarding the small effect found within the SP subdomain, this is probably attributable to the participants having acquired deafness postlingually and therefore having nearly normal speech production to begin with. To find an effect here, a study would likely need to include participants with bilateral peri- or prelingual deafness. The effect of CI provision on HRQoL is important and can be used to help give CI candidates realistic expectations on postoperative results [9].

As shown by many other studies, where the post-test minus the pre-test was used, CI provision leads to a statistically significant improvement in the HRQoL in all subdomains of the NCIQ [8, 9, 15, 19]. These conclusions are in accordance with our results of the observed change in most scores.

Our null hypothesis that the participants’ preoperative t0 scores and the pre-t1 (then-test) scores do not differ had to be rejected. This study is the first one to compare the pre-test and the then-test in the same study population in CI users.

The present study is not without limitations. Future studies would benefit from including more participants than the 17 in the present study, which is a small cohort compared to other studies which used the NCIQ [46,47,48]. As proposed by Sébille et al. [49], different methods should be applied to the same data set to see how they compare with respect to detecting the same type of RS.