This is the first study demonstrating that the patients with planned ICU admission assessed after three months generally have a good memory of their health status as measured by EQ-5D in the period preceding surgery and ICU admission. This memory also appears to remain good after 6 months, because the values of ICC for EQ-5D recalled at 3 months and for EQ-5D recalled at 6 months in comparison with the pre-ICU EQ-5D were both higher than 0.8, which is generally regarded as an excellent concordance . On the other hand, the ICC for the EQ-VAS recalled was just acceptable at 3 months (0.648) and became lower at 6 months in comparison with the pre-ICU level (0.580). The reason for the different behaviour of EQ-5D and EQ-VAS may be strictly mathematical because the former is based on three possible answers (no problems; some/moderate problems; severe/extreme problems) for each of the EQ-5D items, while the latter is on a 101 point scale: the larger the scale parameter, the more spread out the distribution, and the higher the probability of making a different choice.
As far as methodological aspects are concerned, the study hospitals were located in two contiguous and similar Regions of Northern Italy, and the instrument used has been adopted in studies investigating different populations [26–28]. The TTO transformation of patients' EQ-5D was performed using data from a UK population, so those for Italian people may be different. However, considering that the transformation was just used to analyse the statistical agreement - the concordance between the ratings of the same thing and period assessed at different points in time - also different formulas to obtain TTO applied to all EQ-5D ratings would have given the same concordance.
Our findings agree with those of Guadagnoli et al.  who studied 1038 chest pain patients admitted to six hospitals for actual or suspected acute myocardial infarction and found substantial stability over time in response to individual items. The average difference between the scores assessed at the two times was significantly different from zero in only two cases; in both cases, patients reported that they were more functional before admission when asked at 3 months than when asked at the time of hospital stay. Accordingly, the EQ-VAS of our patients showed a slight trend towards increasing over time, suggesting that previous health status may be perceived better as time passes.
The information given by our study may be more useful than expected. In fact, most ICU admissions are unpredictable, so baseline HRQOL is usually measured according to the relatives' opinions . Nevertheless, proxies may not accurately provide baseline measurements due to stress, infrequent contact with the patient, or different perceptions in comparison with the patient [30, 31]. Diaz-Prieto et al  found kappas for patient-proxy concordance ranging from 0.52 for mobility to 0.31 for anxiety/depression, without the effect of the type of patient/proxy relationship, or level of education or admission category (trauma, scheduled or unscheduled surgery, or medical). On the other hand, in the same study EQ-5D VAS scores obtained from patients and proxies correlated much better, with an ICC coefficient of 0.72, which is not so far from that found in the present study (0.648 at three months). Therefore, investigators interested in the before/after comparison of the quality of life of ICU patients may obtain a more reliable assessment of baseline health status interviewing the patients three or six months after discharge than interviewing the relatives.
The multivariate analysis showed that the ability to recall pre-ICU EQ-5D was poorer for general surgery patients at 3 and 6 months, and for urologic surgery patients at 6 months. Possibly, the sequelae of surgery or anti-neoplastic treatments, if required, may affect HRQOL memory in those patients, in comparison with orthopaedic surgery patients.
The similar ability to recall pre-ICU EQ-5D and EQ VAS showed by the patients with ICU LOS ≤ and > 2 days suggests a limited effect of ICU stay on recall and gives strength to our study, despite the significant differences between the two patient groups in the incidence of delirium, hospital LOS and PTSS-14 at 6 months (Table 2). Considering 45 as cut-off for PTSS-14 , only patients with ICU LOS > 2 days had high values (one with 45 at 3 months and 24 at 6 months, and two with 45 and 50, respectively, at 6 months). Accordingly, we cannot exclude that the development of any PTSD symptoms may affect the recall of pre-ICU HRQOL. Interestingly, the chronic use of beta-blockers was associated with better ability to recall pre-ICU EQ-5D, both at 3 and at 6 months. This findings agrees with a recent study showing that a pharmacological blockade of beta-adrenoceptors prevents glucocorticoid-induced memory retrieval deficits in human subjects . A number of studies have examined the influence of giving a β-adrenergic receptor antagonist [34, 35], to try to reduce the incidence of PTSD, however these therapies may be problematic in the critical care population and more research is needed to clarify their role.
As far as study limitations are concerned, our aim was to investigate stability of memory of HRQOL. Therefore, the only population suitable for the on time assessment before ICU admission consisted of patients with planned ICU admission. Consequently, it does not demonstrate that the findings reported are of value for patients with unplanned ICU admissions. Considering that Diaz-Prieto et al  found no relationship between patient-proxy concordance and admission category, we may infer that our findings should be of general value. The exclusion of patients who were not admitted to ICU after surgery despite an admission planned at the time of the anaesthetic visit, allowed a homogeneous sample of patients with the same factors possibly influencing patient memory to be evaluated. In fact, the administration of analgesic and sedatives, which is a common ICU practice, has been demonstrated to influence patient memory of the ICU stay [36, 37]. This practice may also influence the memory of the period preceding ICU admission, so we preferred to study the patients exposed to the same risk factors, that is those really admitted to ICU.