Determinants of ICU memories and the impact on the development and trajectory of post-traumatic stress symptoms: a multicenter longitudinal cohort study

Abstract

Purpose

To identify demographic and clinical determinants of memory formation in intensive care unit (ICU) patients, and determine the relationship between ICU memories and the development and trajectory of post-traumatic stress disorder (PTSD) symptoms.

Methods

Adult patients (n = 426) from two Dutch University ICUs underwent a structured telephone interview using the validated ICU-Memory Tool (ICU-MT) 3 months post-ICU and were assessed for symptoms of PTSD using the Impact of Event Scale (IES-6) at 3 and 12 months post-ICU.

Results

Factual memories without delusional memories were present in 47.7% (n = 203), complete ICU amnesia in 13.8% (n = 59), and delusional memories in 38.5% (n = 164) of patients. Delirium was present in 41% (n = 68) of patients with delusional memories. Using multinomial logistic regression, female sex and number of days with deep sedation were associated with ICU amnesia (aOR 1.99, 95% CI 1.04–3.81, and aOR 1.34, 95% CI 1.09–1.65, respectively), whereas delirium and length of ICU stay were associated with delusional memories (aOR 1.94, 95% CI 1.04–3.61, and aOR 1.11, 95% CI 1.02–1.21, respectively). Of 250 patients assessed at both time points, the prevalence of PTSD symptoms increased significantly over time (4.5 to 10.0%, p < 0.01), driven by a significant increase among those with delusional memories (6.7 to 18.1%, p < 0.01). In a linear mixed-effects model, delusional memories were independently linked to both 3- and 12-month symptoms of PTSD (vs. factuals, adjusted %-difference in mean IES score, 12.9%, 95% CI 2.6–24.1%, and 18.1%, 95% CI 5.2–32.5%, respectively).

Conclusions

Female sex and prolonged deep sedation were associated with complete ICU amnesia, whereas a longer ICU stay and delirium were associated with delusional memories, although these memories were also common in those without delirium. Delusional memories were independently linked to the development and persistence of PTSD symptoms. Targeted interventions to mitigate memory disturbances, both during and after the ICU, may help alleviate the psychological impact of critical illness.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00134-025-08132-4.

Take-home message

In this multicenter longitudinal ICU cohort study, one in seven patients reported amnesia for the ICU, whereas over one-third of patients had delusional memories following discharge. Delusional memories (which were also present in those without recognized delirium) were independently linked to the development and persistence of PTSD symptoms. Mitigating memory disturbances (e.g., delirium management, ICU diaries, guided revisits) might help to alleviate the psychological impact of an ICU admission.

Introduction

The intensive care unit (ICU) can be a highly distressing environment for patients, where various exacerbating stressors such as mechanical ventilation, coma, delirium, and the use of physical restraints can lead to psychological sequelae that may persist for years following ICU discharge [1–3]. Given the trend toward light sedation and increased patient wakefulness in modern ICU practice, more patients may now retain memories of these potentially stressful ICU experiences [4, 5]. The extent to which these experiences are recalled can vary significantly, ranging from complete amnesia for the ICU stay to frightening memories of unreal events, hallucinations, or nightmares, commonly described as delusional memories [6–9].

Delusional memories are prevalent in approximately 30–40% of ICU survivors and are often characterized by a paranoid theme (e.g., poisoning, kidnapping, or staff attempting to assault the patient) [6]. Although these delusional memories often seem bizarre, they are remembered with a substantial amount of detail and remain highly vivid for months or even years following ICU discharge [8, 10]. As a result, these memories may intertwine with or even substitute recollections of factual events that occurred, potentially exacerbating the psychological distress experienced by patients after critical illness [11, 12].

Patients’ recall of their intensive care experiences can be systematically investigated using the validated ICU Memory Tool (ICU-MT), which is specifically designed to assess and categorize different types of ICU memories [13]. Yet, the extent to which demographic and clinical determinants contribute to differences in ICU memory recall is largely unknown. Delirium, a clinical manifestation of acute encephalopathy frequently occurring during critical illness, is often accompanied by hallucinations [14]. Although delusional memories are commonly considered indicative of a prior delirious state [15, 16], multiple studies have reported comparable incidence rates of delusional memories in patients with and without delirium [17, 18]. Furthermore, physical restraints are a known risk factor for development of post-traumatic stress disorder (PTSD) [19], but research exploring the role of physical restraints in the development of delusional memories and subsequent psychological distress is limited [20].

Moreover, the association between different ICU memory types and long-term psychological distress remains poorly understood [15]. The Impact of Event Scale (IES)−6, a well-validated and frequently used instrument to measure symptoms of post-traumatic stress in ICU survivors, can be paired with the ICU-MT to further understand the relationship between ICU memories and psychological distress [21, 22]. Studies investigating the relationship between delusional memories and PTSD have reported conflicting results [20, 23–25]. Furthermore, the role of factual memories remains ambiguous, as some studies suggest a protective effect against delusional memories [6, 26], whereas others have identified factual recall to contribute to PTSD [23, 27]. Notably, the majority of studies had a limited follow-up period of 3–6 months post-ICU [15, 17, 20, 27] and did not investigate the trajectory of PTSD development over time in relation to different memory types. While memories of factual events of the ICU tend to fade over time, delusional memories often persist [6, 10], potentially influencing the occurrence and severity of PTSD symptoms in the long term.

Therefore, the aim of the present study is to identify demographic and clinical determinants of ICU memory formation and to determine the relationship between ICU memories and the development and trajectory of PTSD symptoms.

Materials and methods

Study design and ethics

This multicenter longitudinal cohort study was conducted in Radboud University Medical Center in Nijmegen and University Medical Center Utrecht, the Netherlands, and was approved by the medical ethics committees of both centers (approval numbers 2020-6931 and 20–516, respectively). Each participant or their legal representative provided informed consent during the hospital admission.

Participants

For logistical reasons, participants were included during six 3-month periods, encompassing a total of 18 months from July 2020 up until April 2024. All adult patients (≥ 18 years) admitted to one of the participating ICUs for at least 12 h were initially deemed eligible for inclusion; however, this criterion was amended to a minimum of 48 h after the second 3-month period due to a low overall rate of recollections reported by patients, which posed a risk of misclassification of patients into memory groups. Exclusion criteria were patients admitted for a neurological or neurosurgical reason, patients with pre-existing post-traumatic stress symptoms, or patients unable to speak or understand Dutch.

Data collection

Demographics and clinical data

All demographic data and clinical characteristics were retrieved from the electronic patient files. Severity of illness was assessed using both the Acute Physiology and Chronic Health Evaluation (APACHE) IV score [28] (on the day of ICU admission) and the Sequential Organ Failure Assessment (SOFA) score [29] (daily during the ICU stay). Presence of delirium was assessed two to three times daily by trained nurses using either the Confusion Assessment Method for the ICU (CAM-ICU) [30] or Intensive Care Delirium Screening Checklist (ICDSC) [31]. Level of sedation was assessed using the Richmond Agitation–Sedation Scale (RASS) [32]. Deep sedation was defined as a RASS score of ≤ − 3 and concomitant sedative administration for at least two consecutive nurse shifts (q8h), thereby reducing misclassification due to transient, isolated low RASS scores, such as during intubation.

ICU memories

Memories of the ICU stay were assessed at a median of 3.5 months (IQR 3.0–4.7 months) after ICU discharge through structured telephone interviews by trained study staff using the ICU Memory Tool (ICU-MT) [13]. The ICU-MT is a validated 21-item questionnaire designed to categorize ICU memories into different memory types (Box 1).

In accordance with the manual, a memory type was considered present if the patient recalled at least one corresponding checklist item. To increase the possibility of recall, all questionnaire items were queried, even when patients initially reported no memory of their ICU stay [6]. To minimize confusion with general hospital memories, every patient was instructed to report only ICU-specific memories. If needed, relatives were involved to help the patient distinguish the ICU period from the rest of the hospital stay. Delusional memories are not mutually exclusive with either factual recall or ICU amnesia and may co-occur in the same individual. For the present study, patients were divided into three memory groups: factual memories without delusional memories, complete amnesia for the ICU stay (no factual or delusional memories), or delusional memories (with or without factual memories). The presence of delusional memories was verified through patients’ descriptions of these delusional memories. Of note, the term delusional memories as used in the critical care field differs from its traditional use in psychiatry, where it refers to the sudden onset of false memories unrelated to acute experiences, with strong conviction and resistance to contradictory evidence, rather than misperceived recollections of events [33]. Memories of feelings were not taken into account for this study due to a lack of clinical relevance with regard to our research aim and an inconsistent test–retest reliability [13]. No patients exclusively reported memories of feelings.

Memory types and the corresponding checklist items of the ICU Memory Tool.

Factual memories

Family, alarms, voices, lights, faces, presence of breathing tube, suctioning, darkness, clock, tube in your nose, ward round

Complete amnesia for the ICU

No recall of ICU stay and absence of other memories

Delusional memories

Feeling that people were trying to hurt you, hallucinations, nightmares, dreams

Memories of feelings*

Being uncomfortable, feeling confused, feeling down, feeling anxious/frightened, panic, pain

*Not used in this study

Symptoms of PTSD

Symptoms of PTSD were assessed at both 3 and 12 months post-ICU using the Impact of Event scale (IES)−6 [21], a validated screening tool recommended for ICU patients [22, 34]. The IES-6 consists of six items with a 5-point Likert scale, ranging from “not at all” (score 0) to “extremely” (score 4). A validated threshold mean score of ≥ 1.75 was used to identify the presence of PTSD symptoms [22]. The IES-6 questionnaires were completed either via telephone, online, or on paper, depending on patient preference.

Statistical analysis

Depending on the distribution, continuous variables were presented as mean with standard deviation (SD) or median with first and third quartile (interquartile range, IQR), and categorical variables were presented as proportions. Depending on the number of groups and data distribution, differences in continuous variables were analyzed using independent t-tests, Wilcoxon rank-sum tests, one-way ANOVA or Kruskal–Wallis tests, and categorical data were compared using Chi-squared or Fisher’s exact tests.

To identify demographic and clinical determinants of ICU memory types (factual memories, amnesia and delusional memories), a multinomial logistic regression analysis was performed, with the factual memories group as the reference. Variables were selected a priori, based on previous literature or clinical relevance, and included sex, age, mean SOFA score, mechanical ventilation (yes/no), number of days with deep sedation, delirium (yes/no), use of physical restraints (wrist restraints; yes/no), and length of ICU stay. Absence of multicollinearity (indicated by a variance inflation factor (VIF) < 10) and linearity of the logit were confirmed in the model.

The relationship between ICU memories and the development and trajectory of PTSD symptoms was examined using a linear mixed-effects model, including a random intercept for every patient and all other variables as fixed effects. The included variables were identical to the logistic model described above, but with the addition of the memory groups and a time point variable. The outcome variable mean IES score was log-transformed to normalize the residuals. The model showed no heteroscedasticity or multicollinearity.

To retain all cases in the models, missing SOFA scores and deep sedation days were imputed using the mean and median, respectively. Missing items in the IES-6 questionnaire were imputed with the individual’s mean score if five out of six questions were answered (which was the case for 0 and 11 patients at 3 and 12 months post-ICU, respectively) [35]. Completely missing IES-6 questionnaires at 12 months were not imputed, as mixed-effects models can naturally handle missing data through restricted maximum likelihood estimation, assuming missing is at random [36]. However, to ensure robustness of findings, a sensitivity analysis was performed following imputation of 12-month IES-6 data using a chained random forest-based imputation method [37]. Moreover, consistency of results were checked for all analyses after excluding patients with an ICU stay < 48 h, as the inclusion criterion was amended during patient recruitment. Statistical significance was defined as a two-sided p-value < 0.05. This threshold was adjusted for multiple testing using the Holm method where indicated. All data were analyzed using R version 4.4.1 (R Core Team, 2024), including standard packages along with nnet, lme4, lmertest, dotwhisker, emmeans, and missRanger.

Results

A total of 1658 patients were screened, of whom 1232 were excluded (mostly due to neurologic/neurosurgery admission, in-hospital death or unwillingness to participate), leaving a total of 426 patients (Fig. 1). Overall, 31% (n = 130) of the patients were female, the mean (± SD) age was 62.1 (± 13.1) years, the mean APACHE-IV score was 61.8 (± 21.8), and the median length of ICU stay [IQR] was 3.0 [1.2–6.4] days (Table 1).

Fig. 1.

Study flowchart

Table 1.

Baseline characteristics of the three memory groups

	Overall N = 426	Factual memories w/o delusional memories n = 203	Amnesia for the ICU n = 59	Delusional memories n = 164
Female sex	130 (31%)	54 (27%)	22 (37%)	54 (33%)
Age	62.1 (± 13.1)	63.3 (± 12.5)d	65.2 (± 11.2)d	59.5 (± 14.0)
Admission type
Acute surgery	69 (16%)	22 (11%)	4 (6.8%)	43 (26%)f*,a*
Elective surgery	215 (50%)	123 (61%)d*	38 (64%)d*	54 (33%)
Medical	142 (33%)	58 (29%)	17 (29%)	67 (41%)
APACHE-IV score†	61.8 (± 21.8)	58.0 (± 18.8)	60.4 (± 18.8)	67.6 (± 25.3)f*
SOFA score†	6.4 (± 3.1)	5.7 (± 2.8)	7.0 (± 3.1)f*	6.9 (± 3.2)f*
Mechanical ventilation	372 (87%)	169 (83%)	55 (93%)	148 (90%)
Deep sedation	191 (45%)	58 (29%)	27 (46%)f*	106 (65%)f*,a*
Days with deep sedation†	1.0 [0.0–3.0]	1.0 [0.0–2.0]	1.0 [1.0–3.0]f*	2.0 [1.0–6.0]f*,a
Delirium	110 (26%)	26 (13%)	16 (27%)f	68 (41%)f*
Physically restrained	90 (21%)	24 (12%)	15 (25%)f	51 (31%)f*
ICU length of stay	3.0 [1.2–6.4]	2.0 [0.9–3.7]	1.9 [0.9–3.9]	5.9 [3.0–11.6]f*,a*
Hospital length of stay	12.1 [7.1–24.8]	9.7 [6.0–18.1]	8.2 [6.1–16.9]	19.9 [12.5–33.9]f*,a*

Data are presented as mean (± SD) or median [IQR], depending on distribution. Categorical variables are presented as counts (%)

^f,a,dSignificantly more/higher compared to factual memories (f), amnesia (a), or delusional memories (d), with a * indicating p < 0.01

All p-values were adjusted for multiple testing using the Holm method

^†Missing data for 43 patients (APACHE-IV score), 5 patients (SOFA score), and 2 patients (days with deep sedation); the data shown are unimputed

Factual memories of the ICU stay were present in 83.6% (n = 356) of the patients, of whom 57.0% (n = 203) did not have co-occurring delusional memories. Differences between patients with or without factual memories (either with or without delusions) are shown in Supplemental Table 1. Complete amnesia for the ICU stay was reported by 13.8% (n = 59) of the patients, whereas delusional memories were reported by 38.5% (n = 164). Patients with delusional memories were the youngest, were more often admitted for acute surgery, were more frequently and longer deeply sedated, and had a longer ICU and hospital stay, as compared to those with factual memories or amnesia (Table 1). Delirium was present in 41% (n = 68) of patients with delusional memories.

Patients with delusional but no factual memories were more severely ill, were deeply sedated for more days, and stayed longer in the ICU compared to patients who had both delusional and factual memories (Supplemental Table 2).

Demographic and clinical determinants of different ICU memory types

Both female sex and days with deep sedation were significantly associated with complete amnesia for the ICU stay (adjusted odds ratio, aOR 1.99, 95% CI 1.04–3.81, and aOR 1.34, 95% CI 1.09–1.65, respectively, Fig. 2). On the other hand, delirium and length of ICU stay were associated with the presence of delusional memories (aOR 1.94, 95% CI 1.04–3.61, and aOR 1.11, 95% CI 1.02–1.21, respectively, Fig. 2). The use of physical restraints did not moderate the association between delirium and delusional memories (interaction effect, p = 0.42; data not shown).

Fig. 2.

Demographic and clinical determinants of different ICU memory types (multinomial logistic model).

The reference group is factual memories without delusional memories. Colored lines represent 95% confidence intervals

The prevalence of PTSD symptoms over time between different memory groups

Of the total cohort, 250 patients (58.7%) were alive at 12 months post-ICU and had completed the corresponding PTSD questionnaire, allowing the assessment of PTSD symptoms over time. They differed from non-responders in several aspects (e.g., they were older, more severely ill, more frequently deeply sedated and had a longer ICU length of stay), but had a comparable prevalence of PTSD symptoms at 3 months post-ICU (Supplemental Table 3).

The prevalence of PTSD symptoms increased significantly from 3 to 12 months post-ICU (4.5 to 10.0%, p < 0.01, Fig. 3), which was due to the significant increase among those with delusional memories (6.7 to 18.1%, p < 0.01, Fig. 3). Mean IES scores and subscores between memory groups are shown in Supplemental Table 4. The prevalence of PTSD symptoms grouped by presence of delirium with or without delusional memories is shown in Supplemental Table 5.

Fig. 3.

Prevalence of PTSD symptoms per timepoint and memory group. 

*p < 0.05; **p < 0.01. A mean IES-6 score ≥ 1.75 was used as a cutoff to determine the presence of PTSD symptoms [22]

The relationship between ICU memories and the development and trajectory of PTSD symptoms

Having delusional memories was independently linked to symptoms of PTSD at 3 months post-ICU compared to factual memories (adjusted %-difference in mean IES score, 12.9%, 95% CI 2.6–24.1%, Fig. 4), but not compared to amnesia. At 12 months after ICU discharge, delusional memories remained independently associated with more symptoms of PTSD compared to both factual memories and amnesia for the ICU (adj. %-difference in mean IES score, 18.1%, 95% CI 5.2–32.5%, and 22.8% 95% CI 5.1–43.5%, respectively, Fig. 4). None of the memory groups showed significant changes in PTSD symptom trajectories over time compared to the other memory groups (non-significant interaction effects with time), although individuals with delusional memories showed a trend toward worsening of PTSD symptoms from 3 to 12 months compared to those with amnesia (p = 0.08).

Fig. 4.

The relationship between different ICU memories and the development and trajectory of PTSD symptoms (linear mixed-effects model).

The outcome variable mean IES-6 score was log-transformed to normalize the residuals. Model estimates were exponentiated, allowing them to be interpreted as % changes. Differences between the time points and memory groups were computed using estimated marginal means. The overall difference between 12 vs. 3 months post-ICU was averaged across all three memory groups. The model was adjusted for sex, age, mean SOFA score, mechanical ventilation, days with deep sedation, delirium, use of physical restraints, and length of ICU stay. Colored lines represent 95% confidence intervals

Sensitivity analyses

The results of sensitivity analyses following exclusion of patients with ICU stays shorter than 48 h and following imputation of 12-month IES-6 data were consistent with all primary findings (Supplemental Table 6 and Supplemental Figs. 1–4).

Discussion

In this multicenter longitudinal ICU cohort study, approximately one in seven patients reported complete amnesia for the ICU following discharge, and over one-third of patients had delusional memories. Female sex and days spent deeply sedated were associated with complete ICU amnesia. On the other hand, delirium and the number of days in the ICU were risk factors for delusional memories, although delusional memories frequently occurred in patients without delirium. Having delusional memories was independently linked to both short- and long-term symptoms of PTSD. These results highlight the importance of mitigating memory disturbances in order to alleviate the psychological impact of an ICU admission.

Contrary to prior assumptions and beliefs [38], our findings strengthen the evidence that factual memories of the ICU are not inherently harmful, and the routine use of sedation to suppress these memories, out of concern for psychological outcomes, may not be justified [39]. Instead, a patient-centered, multimodal, symptom-driven approach that prioritizes minimal to no sedation appears more appropriate, as it enables patients to maintain awareness, participate in care, and preserve autonomy [5].

Although delirium was associated with the presence of delusional memories, such memories were also frequently observed in patients without diagnosed delirium. This aligns with previous research suggesting that the presence of delirium is indeed not a prerequisite for the formation of delusional memories [17]. However, the fluctuating nature of delirium, the variability in psychomotor presentations (with the risk of underdiagnosing hypoactive and subsyndromal forms), and the inherent difficulties in assessing sedated patients may at least partly account for this discrepancy [40]. Nonetheless, delirium remains a modifiable factor, and effective prevention and treatment strategies could play a key role in limiting the development of delusional memories and associated psychological sequelae. While the bundled application of non-pharmacological interventions remains the cornerstone of delirium management, the use of dexmedetomidine should be considered in all patients requiring sedation and at risk of delirium, as it is the only evidence-based pharmacological agent to prevent delirium [41]. The effects of antipsychotic treatment—despite no demonstrated benefit on outcomes such as delirium prevalence, duration or mortality—on mitigating delirium-related symptoms such as agitation and hallucinations warrant further investigation [42].

The prevalence of short- and long-term post-traumatic stress symptoms observed in the present study is similar to levels reported in other studies [43]. The increase in PTSD symptom prevalence from 3 months to 1 year post-ICU highlights the need for extended follow-up of patients, even in the absence of PTSD symptoms in the short term (i.e., 3 months), and especially in those with delusional memories. This follow-up could include a guided revisit, allowing patients to see where they were admitted, which could help them process their experiences and further clarify the distinction between real events and distorted perceptions [44]. Additionally, early interventions, such as the use of patient diaries during the ICU stay, may also mitigate the psychological impact of these memories. These diaries can be maintained by both ICU professionals and family members and provide a factual chronology of events, helping patients gain a better understanding of their ICU experiences, which could alleviate the associated psychological burden, also for relatives [45, 46].

Of note, the relationship between delusional memories on the one hand and psychological distress on the other may not be solely unidirectional. While delusional memories are often regarded as a cause of psychological distress, they may instead also serve as manifestations of underlying or subclinical psychological distress [47]. In turn, such distress may also facilitate the persistence of these memories, suggesting a bidirectional rather than a unidirectional pathway.

Strengths of this study are the large sample size, the prospective multicenter design and the long-term follow-up covering multiple time points, allowing the evaluation of PTSD symptom trajectories across different memory groups over time. However, some limitations should be acknowledged. First, 41% of participants did not respond to the 1-year post-ICU questionnaire, despite significant efforts to improve the response rate through targeted email and telephone reminders. Given that responders were generally more severely ill and stayed longer in the ICU, the long-term prevalence of PTSD symptoms might be overestimated. The response rate is, however, largely comparable to other studies examining long-term patient reported outcome measures, particularly in post-ICU populations with inherently lower 1-year survival rates [48]. Second, data on psychological treatment or counseling was not collected for this study, which may have played a confounding role in the trajectory and long-term prevalence of PTSD symptoms. Third, although well-validated for the ICU population, the IES-6 is a screening tool and is not suitable to formally diagnose PTSD, and it may not be sensitive enough to distinguish subtler symptoms of PTSD or specific PTSD symptom clusters [22]. Fourth, the ICU-MT was administered 3 months after the ICU stay, which may have introduced recall bias. To mitigate this, patients were approached for study participation during their ICU admission, and family members were involved to assist in recall if needed (e.g., distinguishing ICU memories from ward memories). Fifth, variation in the administration of IES questionnaires, either by telephone, online or on paper, could have introduced bias. However, the prevalence of PTSD symptoms was comparable across all methods, suggesting minimal influence.

Conclusion

One in seven ICU survivors reported complete ICU amnesia, which was more common among females and those who spent more days deeply sedated. On the other hand, a longer ICU stay and delirium were associated with delusional memories, reported by over one-third of patients, although these were also common in those without delirium. The presence of delusional memories was independently linked to short- and long-term symptoms of post-traumatic stress disorder. Mitigating memory disturbances, both during (non-pharmacologic interventions, A–F bundle) and after the ICU (ICU diary, guided revisits), may help alleviate the psychological burden of critical illness.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

APACHE-IV

Acute Physiology and Chronic Health Evaluation-IV

CAM-ICU

Confusion Assessment Method for the ICU

ICDSC

Intensive Care Delirium Screening Checklist

ICU

Intensive care unit

IES

Impact of Event Scale

IQR

Interquartile range

LOS

Length of stay

OR

Odds ratio

PTSD

Post-traumatic stress disorder

RASS

Richmond Agitation–Sedation Scale

SD

Standard deviation

SOFA

Sequential Organ Failure Assessment

VIF

Variance inflation factor

Funding

None.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Conflicts of interest

None to declare.