Six-month survival and quality of life of intensive care patients with acute kidney injury

We performed a six-month follow-up of 2,901 ICU patients included in the prospective, multi-centre FINNAKI study during the period 1 September 2011 to 1 February 2012. In brief, the study included consecutive emergency ICU patients and elective (postoperative) patients, whose stay exceeded 24 hours. The FINNAKI study in detail has been published previously [3]. In this pre-determined follow-up study we aimed to measure QOL of all six-month survivors. We decided a priori to include ICUs with at least a 70% follow-up rate in the final analysis. Each patient or proxy gave written informed consent for the study. The Ethics Committee of the Department of Surgery in Helsinki University Hospital gave approval for the study.

We measured creatinine (Cr) daily and urine output hourly and defined AKI with the KDIGO guidelines [16]. If baseline creatinine (latest value from the previous year excluding the previous week) was not available, we used the Modification in Diet in Renal Disease (MDRD) equation assuming a glomerular filtration rate of 75 ml/1.73 m²[17]. We collected patient demographics, medical history, severity scores, length-of stay, physiologic data, and the EuroQol (EQ-5D) questionnaire responses from the Finnish Intensive Care Consortium prospective database (Tieto Ltd, Helsinki, Finland) and with a study specific case report form. We obtained data on survival at six months from the Finnish Population Register Centre. We collected physiologic data and screened the patients for the presence of AKI and severe sepsis for five days starting from ICU admission.

The QOL of the included patients at ICU admission and at six months was measured with the EQ-5D [18] questionnaire. The EQ-5D is a standardized, multidimensional instrument found suitable for critically ill patients [19, 20]. The EQ-5D includes five dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) evaluated on a scale of 1 to 3. Calculating a single index score (0 to 1) combines these five dimensions. This value can be used in comparisons between different populations. The questionnaire includes a visual analogue scale (VAS) (values from 0 to 100) for recording the respondent’s self-rated health. According to previous reports a significant change in the EQ-5D index is 0.08, and for the VAS score 7 [21, 22]. The ICU nurse presented the EQ-5D questionnaire to the patient or proxy when collecting the admission data. Data obtained from proxies have been shown to be reliable [23, 24]. At six months, the questionnaire was presented by mail or by telephone.

We present continuous data as medians with interquartile ranges (IQR) and absolute values and percentages with 95% confidence intervals (CI). We used the two-tailed Mann–Whitney U test or the Kruskal-Wallis test for comparison of continuous variables and the chi-square test for categorical variables. We compared the EQ-5D index and reference values using the Wilcoxon signed matched pair test. The analysis for the change in EQ-5D indexes was performed as a paired samples analysis. We explored potential independent factors associated with good QOL (defined as at least equal to that of age- and sex matched controls) at six months. First, we tested plausible factors in univariable models and, second, we entered significant variables (P <0.20) into a multivariable model (logistic regression). Additional file 1 lists all the tested and inserted variables. We considered a P value less than 0.05 as significant unless stated otherwise. We performed all analyses by SPSS version 20 (SPSS, Chicago, Ill., USA).

The study flow chart is presented in Figure 1. Altogether 1,568 patients from ten ICUs with at least a 70% QOL response rate were included in the final analysis. Characteristics of the study patients are presented in Table 1 and a comparison to the whole FINNAKI study cohort is in Additional file 2. The incidence of AKI (95% CI) in this substudy was 635/1,568 (40.5%, 38.0% to 43.0%). Of all patients (95% CI), 280 (17.9%, 15.9% to 19.8%) had stage 1, 119 (7.6%, 6.3 to 8.9%) had stage 2, and 236 (15.1%, 13.2% to 16.9%) had stage 3 AKI. During the first five ICU treatment days 162/1,568 (10.3%, 8.8% to 11.9%) patients received RRT.

The overall six-month mortality (95% CI) was 378/1,568 (24.1%, 21.9% to 26.3%). Of the 635 AKI patients, 224 (35.3%, 95% CI 31.5% to 39.1%) died within six months, as compared to 154/933 (16.5%, 95% CI 14.1% to 18.9%) patients without AKI. The six-month mortality for patients with RRT was 63/162 (38.9%, 95% CI 31.2% to 46.5%).

Of the 1,190 patients who were alive at six months, 959 (80.6%) answered the EQ-5D questionnaire. The median response time was 232 days. The characteristics of the respondents and non-respondents at six months were comparable (data not shown).

Table 2 presents the distribution of the EQ-5D health questionnaire answers in different patient groups at six months. The EQ-5D index and VAS at six months after ICU admission for different patient groups and compared to the age- and sex-matched general population are presented in Table 3.

Of the six-month respondents (N = 959), the admission EQ-5D was available for 774 (80.7%). Of these, 615 (79.4%) were answered by the patient, 139 (18.0%) by a proxy, and in 20 (2.6%) the respondent was unknown. In these 774 patients the mean increases in the EQ-5D index during the six-month follow-up were 0.017 (no AKI) and 0.024 (AKI). The mean difference (95% CI) between the mean changes for non-AKI and AKI patients was 0.007 (-0.314 to 0.045) (P = 0.728). The changes in the EQ-5D index within six months following ICU admission in different patients groups are illustrated in Figure 2.

Of 1,568 study patients, 390 (24.9%) did not respond to the EQ-5D questionnaire on ICU admission. These non-respondents on ICU admission had higher severity scores compared to the respondents (day 1 SOFA score (IQR) 8 (6 to 10), as compared to 7 (5 to 9), and SAPS II score (IQR) 40 (31 to 55), as compared to 35 (26 to 47)).

Of the 959 six-month respondents, 311 (32.4%) achieved a good QOL defined as EQ-5D index which was equal or superior to that of age- and sex matched general population. Of the 327 AKI patients, 96 (29.4%) achieved this good QOL. When exploring potential predictors of good QOL after AKI, we found that (1) the EQ-5D score at admission (odds ratio (OR) (95% CI) 1.042 (1.024 to 1.060)/0.01 points), and (2) lack of hypertension (OR 2.561 (1.141 to 5.750) were independent predictors of a good QOL six months after ICU admission. In non-AKI patients only the admission EQ-5D index (OR 1.039 (1.028 to 1.049)/0.01 points) was an independent predictor of good QOL at six months. The variables tested in univariable models and variables inserted into the multivariable models are listed in Additional file 1. In this study population a longer hospital length-of stay (LOS) was associated with lower QOL (P <0.001) [see Additional file 3].

We reported a six-month mortality of 35.3% for AKI patients, and 16.5% for patients without AKI. The only previous, prospective study found six-month mortality of AKI patients (defined by RIFLE, AKIN or KDIGO criteria) to be 46.5% [13]. In addition, two retrospective, single-center studies have reported six-month mortality rates of 58.5% [10] and 38.0% [25] for AKI patients.

The long-term outcome of RRT patients has been more extensively studied [26]. In our study the six-month mortality for RRT patients was 38.9%. Three other ICU studies, all retrospective, have reported higher six-month mortality rates for RRT patients: 49.4% [27], 59.9% [28], and 74.6% [29].

The QOL of survivors measured six months after ICU admission did not differ between patients who had developed AKI and those who had not. Moreover, the QOL did not change during this follow-up period for either group. However, the QOL of ICU patients was already significantly lower than that of the age-and sex-matched general population at the time of ICU admission and remained so during the follow-up period (Table 3). Recently, comparable results obtained with a different QOL questionnaire were reported by Hofhuis et al. [13]. In contrast, in another study from 2009, AKI patients had a lower QOL six months after surgery compared to patients without AKI [25].

Other QOL studies in AKI patients have only included RRT patients, and all have reported impaired health compared to controls [5, 27, 30‐34], in concordance with our results. A randomised, controlled trial comparing different intensities of RRT reported low QOL scores for RRT patients at 60 days [6]. Compared to our study, however, that study had a different QOL measure (Health Utilities Index (HUI) versus EQ-5D), follow-up time (60 days versus 180 days), patient population (only RRT versus all AKI patients) and reported no admission QOL data or VAS scores. Two Finnish studies with the same QOL questionnaire used in our study (EQ-5D) also reported that the QOL of RRT patients was lower compared to the matched general population [5, 27].

RRT patients, despite their impaired QOL, have self-rated their quality of life as equal to that of the age- and sex-matched general population [5, 27] or stated that they would choose to undergo the same treatment again [30, 32]. It has been suggested that surviving critical illness affects how people value their life and they are, therefore, content with health, which by objective measures is lower than that of the general population. In our study, AKI patients were as content with their QOL as the general population, and ICU patients without AKI evaluated their QOL to be even better compared to the general population by VAS. RRT patients, however, reported significantly lower VAS scores compared to the general population (Table 3).

The QOL of AKI and non-AKI patients was significantly lower than that of the general population already at the time of ICU admission which is in concordance with previous data [35]. In addition, the change in QOL in survivors of critical illness was minimal. This indicates that critical illness per se may not impact the QOL, but patients who get critically ill have existing poor health. Furthermore, patients who died during critical illness had a significantly lower EQ-5D at ICU admission compared to patients who survived.

One third of the survivors and one third of the AKI patients had an equal or superior QOL compared to the age- and sex-matched general population at six months. We found only two independent predictors of good QOL at six months after AKI: a higher EQ-5D index at ICU admission was associated with a higher probability of a good QOL and lack of hypertension as a chronic condition added the probability of a good QOL. Surprisingly, age, chronic conditions other than hypertension, or events prior to ICU admission were not associated with the QOL outcome at six months in AKI patients. For comparison, in non-AKI patients only the EQ-5D index at admission was an independent predictor of a good QOL at six months in this study population. According to a systematic review from 2005, older age and increasing severity of illness may be associated with poorer outcome in some QOL dimensions in EQ-5D (physical function and general health perception) [35]. In our population, the patients with the highest tertile of LOS had the lowest QOL at six months in agreement with one previous study [6]. The systematic review from 2005, however, found no association between ICU LOS and QOL [35]. Furthermore, hospital LOS is a non-normally distributed factor that is also affected by factors unrelated to the patient. These findings reflect the difficulties in finding individual markers that would help us predict which patients will have a favorable outcome in the form of a good QOL after ICU treatment.

Our study has some limitations. First, although the mortality data at six months were anticipated to be complete due to the national registry, we expected that QOL data would be incomplete in some sites. Thus, to provide a reliable analysis we decided to include only those sites with at least a 70% QOL response rate. This decision resulted in the exclusion of 7 of 17 sites. Second, only 80% of those patients who survived in the included ten sites answered the EQ-5D questionnaire at six months. Finally, the pre-admission QOL was not available for 19% of six-month respondents. Thus, our analysis of QOL change between these two time-points was not complete. However, of the patients without AKI, one could assume that those patients who responded were slightly less ill than those who did not respond. In fact, that was the case and, thus, our finding of a non-significant difference of mean changes of QOL in critically ill patients with AKI and without AKI seems to be more reliable.

An obvious strength of our study is the prospective, multi-centre design and consecutive inclusion of a large number of patients both with and without AKI in the same ICUs and the availability of admission QOL data. In addition, we aimed to explore carefully the inevitable selection bias commonly seen in QOL studies in the critical care setting.