3.1 Study design

We performed this systematic review and meta-analysis in accordance with the preferred reporting items for systematic reviews and meta-analyses (PRISMA) methodology.²² The protocol was registered with PROSPERO (#CRD42020198973). The present study was performed to demonstrate the clinical effects of rehabilitation or the ICU diary on ICU patients. The rehabilitation and ICU diaries examined in this study are described in Tables S1–S3.

We performed a comprehensive literature search of PubMed (Table S4), CHINAHL (Table S5), all Ovid journals (Table S6), and CENTRAL (Table S7) for relevant articles published between 1 January 1985, and 19 October 2022. Before the search was performed, the search protocol was augmented and assessed using the Peer Review of Electronic Search Strategies (PRESS) guidelines by an experienced librarian.²³

Authors independently reviewed the title and abstract of each of the articles identified in the literature search to select potential articles. Regarding patients, populations, and issues, inclusion criteria were adults (older than 18 years) admitted to the ICU with critical illnesses whose cases were described in English-language publications or unpublished literature dated after 1 January 1985. Selected studies were limited to RCTs in order to ensure an appropriate level of evidence. Regarding interventions, the inclusion criterion was rehabilitation or the ICU diary for patients who had stayed in or were discharged from the ICU.

Exclusion criteria were duplicate studies or studies involving the same database or population and patients younger than 17 years. The full text and references of each article or abstract that passed the initial screening were analysed to identify any articles that had been missed. Specifically, the full text of each article selected during the initial screening and the references listed within them were read by each reviewer, and any relevant studies were selected according to the predetermined inclusion and exclusion criteria. The four abovementioned authors independently reviewed each article for final inclusion at the full-text screening stage, and group consensus was used to resolve conflicts.

3.2 Data collection and processing

Authors independently analysed each of the included studies to extract information, which was entered into an Excel spreadsheet. The information extracted from each article included a description of the research outline (title, first author, journal name, and the year of publication), the research design (purpose, research period, subject selection criteria and exclusion criteria, age, admitting diagnosis, severity, ICU stay period, total number of subjects recruited, and the total number of subjects analysed), the interventions performed (contents, treatments employed in the control group, period, and endpoints), and the study outcomes (results, sample size, and p-values). Group consensus was used to resolve any conflicts regarding the extracted data.

The abovementioned authors independently assessed the risk of bias for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions and the Cochrane Effective Practice and Organization of Care (EPOC) guidance. We assessed random sequence generation, allocation concealment, the blinding of participants and personnel, the blinding of outcome assessments, incomplete outcome data, selective reporting, and other sources of bias. We judged each potential source of bias as high, low, or unclear and justified our judgement. We used the EPOC risk of bias guidance to help reach our judgements. Other authors settled any disagreements between the independent authors. Publication bias was not evaluated because the number of studies in each subgroup was 10 or less.

3.3 Outcome measures

The primary outcome was the HADS score. The secondary outcomes included SF-36, EQ-5D-5L, and IES-R scores. HADS is a measurement scale for anxiety and depression that is widely used worldwide. It consists of 14 items, including seven items each for anxiety and depression, and each item is measured on a 4-point scale (range: 0–21).¹⁸ The SF-36 measures health-related QOL. It is used to produce a physical component summary (PCS, range: 0–100), which measures physical aspects, including physical function, physical role functioning, body pain, and general health, and a mental component summary (MCS, range: 0–100), which measures mental aspects, including vitality, social functioning, emotional role functioning, and mental health.¹⁹ The score for each domain ranges between 0 and 100, with an average of 50 and a standard deviation (SD) of 10. The EQ-5D-5L measures QOL and is used to calculate quality-adjusted life-year values. It consists of five questions, each of which is measured on a 5-point scale, and, thus, may yield 3125 combinations of QOL scores (range: 0–1.0).²⁰ The IES-R is a self-completed questionnaire that measures the severity of PTSD symptoms. It consists of 22 items, including eight related to intrusion symptoms, eight related to avoidance symptoms, and six related to hyperarousal symptoms, each of which is assessed using a 5-point scale (range: 0–88).²¹

We excluded the State–Trait Anxiety Inventory, Beck Depression Inventory-II, PTSD Symptom Scale, WHO-QOL, and Functional Assessment of Cancer Therapy because of an insufficient number of outcomes.

3.4 Primary data analysis

The pooled effects of rehabilitation or the ICU diary were assessed using a random-effects model and the inverse-variance method. Mean differences (MD) and 95% confidence intervals (CIs) were calculated. Estimates of the mean and SD were calculated for data given as medians and interquartile ranges.²⁴ Results are displayed graphically on forest plots. The observed heterogeneity for the summary and subgroup analyses was measured with the I² statistic.²⁵ Data synthesis and statistical analyses were performed with Review Manager (version 5.4; The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark). Funnel plots were created and visually inspected to assess the presence/absence of publication bias.

3.5 Intervention

We analysed studies that provided rehabilitation or an ICU diary. Rehabilitation was defined as exercise, education, the provision of information, and psychosocial support.²⁶ We focused on exercise, which involves muscular and physical activities to increase cardio-respiratory work and practices to improve performance.²⁶ Rehabilitation programmes in the present study included the following: turning side-to-side on the bed, isometric and resistance exercises, sitting on the edge of the bed, moving from the bed to a chair, cycling ergometry, and walking (Tables S1 and S2). Because clinical outcomes are affected by the timing of the initiation of rehabilitation,²⁷ we divided patients into subgroups that started rehabilitation during the ICU stay or after discharge from the ICU. The ICU diary is generally completed on behalf of patients to provide events recorded throughout their admission to the ICU (Table S3).²⁸

4.1 Search results

The PRISMA flow diagram for rehabilitation is shown in Figure 1. The database search for relevant studies on rehabilitation resulted in 12 135 studies being identified, with an additional eight studies. We excluded 4831 studies because of duplication, and the titles and abstracts of the remaining 7312 studies were screened. As a result, we excluded 7287 unrelated studies. We reviewed the remaining 25 full-text articles to further evaluate their eligibility. We excluded eight studies for which data were unavailable from the body or appendices,^29-36 three in which interventions were not categorized as rehabilitation,^37-39 and two were excluded for other reasons.^{40, 41} Therefore, we included 12 studies in our meta-analysis.^{12, 13, 42-51}

The PRISMA flow diagram for the ICU diary is shown in Figure 1b. The database search for relevant studies on the ICU diary resulted in 11 995 studies being identified, with an additional three studies. We excluded 4365 studies because of duplication, and the titles and abstracts of the remaining 7633 studies were screened. As a result, we excluded 7605 unrelated studies. We reviewed the remaining 28 full-text articles to further evaluate their eligibility. We excluded 7 studies for which data were unavailable from the body or appendices,^52-58 six that had non-RCT designs,^59-64 eight in which interventions were not categorized as the ICU diary,^65-72 and two were excluded for other reasons.^{73, 74} Therefore, we included five studies in our meta-analysis.^{14, 75-78}

4.2 Study characteristics

The demographics of the populations in rehabilitation studies are shown in Table 1,^{12, 13, 42-51} and those of the populations in ICU diary studies are shown in Table 2.^{14, 75-78} Details on rehabilitation programmes and the ICU diary are shown in Tables S2 and S3. The follow-up period lasted ≤12 months in all studies.

A summary of the risk of bias in each study is shown in Figure S1 and a graph of the risk of bias in Figure S1. The randomization methods employed in two studies were unknown.^{49, 50} The blinding of subjects and researchers was categorized as high risk for all studies because of the nature of the study protocols. Only three studies were categorized as low risk for incomplete outcome data because of drop-outs associated with critical illness.^{43, 51, 75}

4.3 Meta-analysis findings

A forest plot of HADS scores comparing standard care (SC) with rehabilitation is shown in Figure S1. HADS-anxiety and HADS-depression scores did not significantly differ between the SC and rehabilitation groups. The funnel plots of HADS-anxiety and HADS-depression for the rehabilitation studies exhibited asymmetric configurations (Figure S1).

A forest plot of SF-36 scores comparing SC with rehabilitation is shown in Figure 2. Significant differences in SF-36 PCS [MD = 2.46, 95%CI: (1.11 to 3.81), p = 0.0004: Figure 2a] and MCS scores [MD = 2.94, 95%CI: (1.23 to 4.65), p = 0.0008: Figure 2b] were detected between the SC and rehabilitation groups. Rehabilitation that started during the patient's ICU stay did not have any significant effect on the PCS or MCS score. On the other hand, rehabilitation that started after the patient was discharged from the ICU led to significant improvements in PCS [MD = 3.31, 95%CI: (1.33 to 5.28), p = 0.001: Figure 2a] and MCS [MD = 4.31, 95%CI: (1.48 to 7.14), p = 0.003: Figure 2b] scores. Funnel plots of SF-36 PCS and MCS scores for rehabilitation studies exhibited asymmetric configurations (Figure S1).

A forest plot of EQ-5D-5L scores comparing SC with rehabilitation is shown in Figure S1. The mean EQ-5D-5L score did not significantly differ between the SC and rehabilitation groups. The funnel plot of EQ-5D-5L scores exhibited an asymmetric configuration (Figure S1).

A forest plot of HADS scores comparing SC with the ICU diary is shown in Figures 3a,b. The mean HADS-anxiety score significantly differed between the SC and ICU diary groups [MD = 0.96, 95%CI: (0.21 to 1.71), p = 0.01: Figure 3a], whereas the mean HADS-depression score did not. The funnel plots of HADS-anxiety and HADS-depression for ICU diary studies exhibited asymmetric configurations (Figure S1).

A forest plot of IES-R scores comparing SC with the ICU diary is shown in Figure 3c. The mean IES-R score did not significantly differ between the SC and ICU diary groups. The funnel plot of IES-R scores exhibited an asymmetric configuration (Figure S1).

5.1 Anxiety and depression

Rehabilitation did not ameliorate anxiety or depression in ICU patients. Rehabilitation is designed to maintain or improve physical functions. Exercise alone cannot improve psychosocial health unless additional interventions involving either psychosocial or educational programmes are provided.⁴⁷ Therefore, additional specific methods other than exercise need to be considered in order to ameliorate anxiety or depression in patients.

The ICU diary effectively ameliorated anxiety, which is consistent with previous findings.⁷⁶ The ICU diary helped critically ill patients reinforce their factual memory and distinguish between hallucinations and delusions.¹⁴ Therefore, it is an effective tool for reducing anxiety in ICU patients. On the other hand, the present results showed that the current ICU diary did not ameliorate depression in ICU patients. Depression differs from anxiety because of the lack of positive affect, temporal orientation, rhythmicity, rumination, and disinvestment in the self.⁷⁹ This psychological gap may be the reason for the difference in the clinical effectiveness of the ICU diary against anxiety and depression. The ICU diary is helpful for organizing memories to reduce anxiety, but not for ameliorating depression.

5.2 PTSD

The ICU diary did not effectively improve the IES-R scores of critically ill patients; it is intended to prevent PTSD. PTSD symptoms are negative flashbacks, nightmares, and the avoidance of external reminders.⁸⁰ The memories recalled by patients using the ICU diary may be negative flashbacks; nevertheless, they may improve anxiety. Because avoidance is a major symptom of PTSD, patients may avoid using the ICU diary to prevent being reminded of traumatic memories.⁷⁵ Therefore, another programme needs to be developed that supports PTSD patients with mental issues. The ICU diary may ameliorate anxiety in critically ill patients.

5.3 QOL according to the SF-36

Rehabilitation significantly improved the SF-36 scores of ICU patients. It improved both PCS and MCS. Shelly et al. suggested that exercises increase the release of endorphins, promote circulation, and improve reactivity to stress.¹² In addition, the physical functional status covaries with multiple mental health-related variables.⁸¹ A previous study reported a moderate to strong correlation between PCS and MCS.⁸² In terms of the SF-36, rehabilitation may improve patients' MCS scores through synergistic effects.

The present results showed that the initiation of rehabilitation during the ICU stay did not significantly affect the SF-36 PCS or MCS scores of ICU patients. On the other hand, rehabilitation effectively improved QOL when it was initiated after the patient was discharged from the ICU. A potential reason for the lack of effect of early rehabilitation initiated during the ICU stay is that an acute inflammatory response suppressed physical responses and muscle rearrangement under critical conditions.⁸³ These scores are also affected by decreased daily activity levels and a worse nutritional status. Because the condition of critically ill patients may rapidly change, the provision of rehabilitation at an appropriate intensity and time is important, and it also needs to be personalized for inflammatory conditions.^{12, 84} The present study shows the need for rehabilitation programmes tailored to each patient's condition.

The minimum mean ± SD of the MCS score in the present study was 41.3 ± 10.6, while the maximum score was 51.9 ± 12.8.^{48, 51} This range may be too narrow to allow any effects of the interventions to be proven. The SF-36 MCS score ranges between 0 and 100. A detection bias needs to be considered when we deal with patient reported outcome (PRO). Patients who stay in the ICU often have difficulty responding to PRO on their own because of disease or medication. In these cases, the nurse has to read the questionnaire and respond on behalf of patients. As a result, the time to complete the questionnaires increases to 15–20 min, even though it typically takes only 5 minutes. A detection bias may arise on the SF-36 because of the large number of questions when it is used for critically ill patients. The timing of use of the SF-36 needs to be considered; however, it is a robust tool that is recommended to assess the QOL of critically ill patients.

5.4 QOL according to the EQ-5D-5L

The EQ-5D-5L was not adequate for examining the effects of rehabilitation, even though it is useful for evaluating QOL in the general population. There are several explanations for why the EQ-5D-5L score did not appear to be useful for the purposes of this study. Because it focuses on physical indicators,²¹ it may be unsuitable for measuring QOL in critically ill patients. Furthermore, the sensitivity of the EQ-5D-5L in critically ill patients may not be sufficient because of the ceiling effects of the utility score.⁸⁵ Therefore, it was challenging to detect any effect of the examined support programmes using the EQ-5D-5L. The EQ-5D-5L may be unsuitable for evaluating the QOL of critically ill patients because of its nature and low sensitivity. Specific QOL evaluation tools need to be established for critically ill patients in the future.