Background

The number of patients admitted to the intensive care unit (ICU) is increasing, and more than 80% of patients survive ICU care [1, 2]. Accordingly, increasing attention is being given to the long-term outcome of ICU survivors. Previous studies have shown that ICU survivors are affected by a wide range of physical, cognitive, and psychiatric issues [3, 4]. These phenomena, also referred to as “post-intensive care syndrome (PICS)”, decrease the patients’ quality of life and deteriorate their ability to return to the workforce [1, 5]. Thus, the Society of Critical Care Medicine released the ABCDEF bundle, which encourages assessment and management of pain, agitation, and delirium, implementation of early mobilization, and family engagement, along with a recommended strategy to mitigate the risk of PICS [6, 7].

Cognitive impairment and psychiatric problems worsen the quality of life of patients who have recovered from a critical illness, and they have become important public health issues [8,9,10,11]. Hopkins and colleagues reported that 45% of ICU survivors had cognitive sequelae and 29% had symptoms of depression and anxiety at 1 year [9]. Sivanathan and colleagues found that ICU patients had a higher rate of diagnoses related to new psychiatric problems in the year after discharge than other hospitalized patients [11]. Therefore, various efforts, including early mobilization, minimization of sedation, and emotional and psychological support, have been made to prevent cognitive impairment and psychiatric problems in patients receiving ICU care [12].

Delirium is a frequent disorder encountered in ICU patients and is associated with substantially increased morbidity and mortality after ICU care [13,14,15]. Despite various efforts, a recent meta-analysis reported that the prevalence of delirium in ICU patients remains high, at approximately 31% [16]. Several studies have demonstrated the impact of delirium on cognitive impairment and psychiatric problems [4, 17,18,19,20]. However, these previous studies were small prospective cohort studies and limited to showing the impact of delirium on detailed post-ICU problems and medical costs. Therefore, we conducted this nationwide cohort study to evaluate the association between the presence of delirium during hospitalization involving ICU care and post-discharge cognitive impairment or psychiatric problems.

Methods

Study population

We analyzed a retrospective cohort from the Korean ICU National Data (KIND) Study [21] based on the Health Insurance Review and Assessment (HIRA) database from the Korean Ministry of Health. The KIND database include all ICU admissions in Korea. From the KIND database, we selected initial ICU admissions in adults (aged over 18 years) without a history of clinic visits for psychiatric disorders and psychiatric medications at admission between January 1, 2007, and May 31, 2016 (n = 345,394). Among the participants, we excluded patients who died before discharge (n = 32,902) and those who had no clinical records after discharge (n = 6481). The final sample size was 306,011 (198,491 men and 107,520 women) (Fig. 1).

Fig. 1
figure 1

Flow diagram

Full size image

The study was reviewed by the Institutional Review Board (IRB) of Samsung Medical Center (IRB protocol 2015-11-17) and was exempted from the informed consent requirement, because we used only de-identified previously collected administrative data.

Data collection and clinical outcomes

Delirium was defined as the presence of psychiatric disease codes (F-code) or a code for antipsychotic or anxiolytic medications in a prescription claim within an admission. We also conducted a sensitivity analysis that used a more specific definition of delirium: the presence of the code for delirium (F05.x except F05.1) instead of all psychiatric disease codes. Comorbidities were defined as a 1-year history of any disease before admission to the ICU and summarized using the Charlson’s comorbidity index [22].

The primary outcome was cognitive impairment and psychiatric problems within 2 years of discharge. Cognitive impairment was defined as the presence of dementia (F00, F01, F03, F03, G30, and G31) or cognitive dysfunction (R41) codes or prescriptions of anti-dementia agents such as cholinesterase inhibitors (donepezil, rivastigmine, and galantamine), memantine, choline alfoscerate, oxiracetam, nicergoline, acetyl l-carnitine, and piracetam for at least 28 days. Psychological impairment was defined as the presence of disease codes for depression (F32, F33, and F34.1), anxiety (F40 and F41), somatoform/conversion disorder (F43), substance abuse disorder (F10–F19), and sleep disorder (F51 and G47) or prescription of antidepressants or antipsychotic agents for at least 28 days. The secondary outcomes were readmissions to the hospital, readmissions to the ICU, and medical costs during hospitalization and after discharge.

Data on underlying diseases and comorbidities, procedures, prescriptions, and patient demographics were obtained from claims data and defined using International Classification of Diseases 10th revision codes [23]. Comorbidities were defined as a 1-year history of any disease before admission to the ICU [24, 25].

Admissions were classified as medical or surgical. Medical admissions were defined as admissions to one of the following departments: pediatrics, internal medicine, neurology, neuropsychiatry, dermatology, rehabilitation medicine, general medicine, radiology, family medicine, and emergency medicine. Surgical admissions were defined as admissions to one of the following departments: general surgery, orthopedic surgery, neurosurgery, thoracic and cardiovascular surgery, plastic surgery, ophthalmology, otorhinolaryngology, urology, oral surgery, anesthesiology, and obstetrics and gynecology.

The quality of ICU care was determined using the codes of nurse staffing grades (AJ11–AJ19, AJ21–AJ9, and AJ31–AJ39). Nurse staffing grades (ratio of number of beds to number of nurses) were categorized into nine grades (grades 1–9). Ratios of number of beds to number of nurses of < 0.5 and ≥ 2.5 indicated grade 1 (best) and grade 9 (worst), respectively.

The procedures of interest were mechanical ventilation (Korean National Health Insurance procedure codes: M5857, M5858, and M5860), extracorporeal membrane oxygenation (ECMO; Korean National Health Insurance procedure codes: O1901–O1904), and continuous renal replacement therapy (CRRT; Korean National Health Insurance procedure codes: O7031–O7034 and O7051–O7054). The use of vasopressor drugs such as dobutamine, dopamine, epinephrine, and norepinephrine for > 2 days was determined according to the Korean Drug and Anatomical Therapeutic Chemical Codes (148201BIJ, 38900BIJ, 148701BIJ, 148702BIJ, 429500BIJ, 152601BIJ, and 203101BIJ) [25]. Advanced interventions were defined as procedures, such as mechanical ventilation, ECMO, CRRT, and administration of vasopressor drugs.

In-hospital cost was defined as the sum of the insurance payment costs incurred during hospital admission, and post-ICU cost was defined as the sum of payments per visit during the 2 years after discharge. To obtain the cost related to cognitive impairment or psychiatric problems, we summed all costs related to codes with cognitive impairment or psychiatric problems as the principal reason during 2 years after discharge.

Statistical analyses

Data are presented as means with standard deviations or medians and interquartile ranges for continuous variables, and as numbers and percentages for categorical variables. Demographic characteristics were compared between the delirium and non-delirium groups using the Mann–Whitney U test for continuous variables and Pearson’s Chi-square test for categorical variables. Patients were followed up until the development of a study endpoint, readmission, 2 years after discharge, or available last clinic visit upto the end of the study period (May 31, 2016). Cumulative incidences and hazard rates were estimated using the Kaplan–Meier method. We calculated the hazard ratios (HRs) with 95% confidence intervals (CIs) for developing each clinical outcome using a proportional hazards regression model. For the cost analysis, we used log-transformed cost as the outcome (because the costs were markedly right-skewed) and estimated the ratio of cost (with 95% CI). Post ICU cost were obtained by the incremental healthcare costs after discharging ICU. With consideration the different follow-up period for each patient, we calculated the PPPM (per patient per month) cost. Each patient’s total cost during the follow-up was divided by the corresponding patient’s follow-up period. All analyses were adjusted for potential confounding factors, such as age, sex, comorbidities, medical beneficiaries, region, mechanical ventilation, ECMO, CRRT, vasopressor drugs, quality of ICU, and length of ICU stay more than 7 days. Statistical significance was set at p < 0.05. Statistical analyses were performed using SAS® Visual Analytics (SAS Institute Inc., USA).

Results

Of the all 306,011 patients, the proportion of those with delirium was 55.0% (n = 168,190). There were 64.9% men, and the mean patient age was 58.3 (15.1) years (Table 1). Patients with delirium were more likely to undergo advanced interventions during hospital admission and showed increased length of ICU stay (3 [1–6] versus 2 [0–4] days, p < 0.001) and hospital day (16 [9–30] versus 8 [4–14] days, p < 0.001) (Table 1). Moreover, patients with delirium were more likely to be admitted to the medical department, than those without delirium, although the reasons for admission were not significantly different between the two groups (Additional file 1: Table S1). The Charlson’s comorbidity index was higher in patients with delirium than in those without delirium (2 [0–4] versus 2 [0–3], p < 0.001). Furthermore, congestive heart failure (8.3% versus 7.1%, p < 0.001), chronic pulmonary disease (26.5% versus 24.9%, p < 0.001), connective tissue disease (3.1% versus 2.9%, p = 0.002), liver disease (23.6% versus 22.4%, p < 0.001), paraplegia and hemiplegia (1.3% versus 1.1%, p < 0.001), and cancer (21.0% versus 19.8%, p < 0.001) were more common in patients with delirium than in those without delirium.

Table 1 Characteristics of study participants
Full size table

The incidence rates of cognitive impairment and psychiatric problems were higher in patients with delirium than in those without delirium (Table 2). The HR for cognitive impairment and psychiatric problems was 1.25 (95% CI 1.07–1.48) and 1.80 (95% CI 1.67–1.94) in patients with delirium compared with those without delirium, respectively. After adjusting for potential confounding factors, the HRs for cognitive impairment and psychiatric problems were also significant at 1.17 (95% CI 1.05–1.29) and 1.78 (95% CI 1.67–1.90) in patients with delirium compared with those without delirium, respectively. In the sensitivity analysis, which used a specific definition of delirium, the HRs for cognitive impairment and psychiatric problems were 1.45 (95% CI 1.40–1.51) and 1.40 (95% CI 1.31–1.51) in patients with delirium compared with those without delirium, respectively (Additional file 1: Table S2). After adjusting for confounding factors, the adjusted HR for cognitive impairment remained significant (1.10; 95% CI 1.07–1.13). In patients who needed advanced interventions, the HR for cognitive impairment was not different (1.08; 95% CI 0.83–1.40, p = 0.59), whereas the HR for psychiatric impairment presented a significant difference (1.70; 95% CI 1.59–1.82) (Table 2). The cumulative incidences of cognitive impairment, psychiatric problems, and hospital readmission were significantly higher in patients with delirium than compared in those without delirium (Fig. 2).

Table 2 Hazard ratio (95% CI) for cognitive impairment and psychiatric problems within 2 years of discharge in patients with delirium versus patients without delirium
Full size table
Fig. 2
figure 2

Cumulative incidences of outcomes in patients with and without delirium. a Cognitive impairment, b psychiatric problems, and c readmission

Full size image

In patients with delirium, the most common psychiatric impairment was sleep disorder (148.7/1000 person-years), followed by depression (133.3/1000 person-years) and anxiety (123.5/1000 person-years, Additional file 1: Table S3). In patients without delirium, the most common psychiatric impairment was anxiety (86.1/1000 person-years), followed by sleep disorder (77.9/1000 person-years) and depression (60.2/1000 person-years). After adjusting for potential confounding factors, the HRs for sleep disorder, depression, anxiety, substance abuse disorder, and stress reaction/conversion disorder were significantly increased in patients with delirium compared with those without delirium.

Most common disorder across all combinations of cognitive impairment and detailed psychiatric problems was cognitive impairment alone, followed by sleep disorder alone and anxiety alone. Cognitive impairment and depression frequently co-occurred (n = 3233). Cognitive impairment also co-occurred with sleep disorder (n = 1791) and anxiety (n = 1683) (Fig. 3).

Fig. 3
figure 3

UpSet plot of the combinations of cognitive impairment and detailed psychiatric problems. The vertical bars reflect the number of participants who experienced a specific combination of cognitive impairment and detailed psychiatric problems, identified by dots below each bar. The horizontal bars reflect the number of participants who experienced each cognitive impairment and detailed psychiatric problems across all combinations of cognitive impairment and detailed psychiatric problems

Full size image

With respect to cost outcomes, an average of $7693 was spent in the ICU during hospitalization. The cost was $9429 and $5732 in patients with and without delirium, respectively. Patients with delirium spent 1.26 times (95% CI 1.26–1.27) more than patients without delirium. In terms of post-ICU cost, the average cost was $5268. Patients with and without delirium spent $6377 and $5082, respectively, and patients with delirium spent 1.15 times (95% CI 1.14–1.17) more than patients without delirium. Among the post-ICU expenses, the cost related to cognitive impairment or psychiatric problems related was $141 and $75 in patients with and without delirium, respectively. Patients with delirium also had higher cost related to cognitive and psychiatric impairments after ICU admission than those without delirium (Table 3). When we compare to the PPPM as consider the follow-up duration, the results were also similar with total cost (Table 3).

Table 3 Ratio (95% CI) of medical costs in patients with delirium compared with patients without delirium
Full size table

Discussion

In this study, we investigated the association between the presence of delirium during hospitalization involving ICU care and post-discharge cognitive impairment or psychiatric problems. The key strength of this study is the first national cohort study to show the association between delirium and cognitive impairment or psychiatric problems seen in a small prospective cohort from previous studies. In addition, this study included a long post-discharge period and demonstrated the impact of delirium on post-ICU medical costs.

Delirium increases the risk of cognitive impairment

In this nationwide cohort study, we found that the incidence of cognitive impairment was 15.5% within 2 years after hospital discharge in ICU survivors. Similar to previous reports, this study showed that patients with delirium had an increased risk of cognitive impairment compared with those without delirium (HR, 1.17; 95% CI 1.05–1.29). Previous studies have shown that delirium is a significant independent risk factor for long-term cognitive impairment [4, 26,27,28]. Pandharipande and colleagues conducted a prospective observational study to identify potentially modifiable risk factors of long-term cognitive impairment in mixed ICU patients. They observed that a longer duration of delirium during hospitalization was associated with worse cognition at 3 and 12 months [4]. Girard and colleagues also found that delirium was associated with long-term cognitive impairment according to several clinical phenotypes (i.e., hypoxic, septic, sedative-associated, metabolic, and unclassified) at 3 and 12 months of follow-up. Their study showed that sedation, hypoxia, and sepsis-related delirium increased the risk of long-term cognitive impairment [27]. Considering that cognitive impairment is a common and severe complication and affects the long-term quality of life of ICU survivors [4, 12], efforts to prevent delirium may be warranted to reduce the risk of long-term cognitive impairment.

Delirium increases the risk of psychiatric disorder

In previous studies, comorbid conditions, benzodiazepine prescription, and preexisting psychiatric disorders were associated with an increased risk of psychiatric problems experienced by ICU survivors [29, 30]. The association between delirium and psychiatric problems is controversial. Jacson and colleagues conducted a prospective multicenter cohort study with 821 patients and reported that delirium was not consistently associated with psychiatric problems [2]. In contrast, some studies have shown that delirium is associated with psychiatric problems [31,32,33]. In the current study, we found that the presence of delirium during hospitalization involving ICU care increased the risk of psychiatric problems within 2 years of discharge (HR, 1.78; 95% CI 1.67–1.90). In detail, the risk of substance abuse disorder, sleep disorder, or depression was significantly increased in patients who experienced delirium compared with those who did not. We analyzed patients who were hospitalized with ICU admission, and the results suggested that delirium in the ICU and immediately after the ICU stay may be important with respect to long-term psychiatric problems. Further studies are needed to determine the relationship between the timing of delirium and the development of psychiatric problems.

Association between cognitive impairment and psychiatric problems

Physical problems, cognitive impairment, and psychiatric problems might be distinct domains of critical illness rather than a component of a single unifying syndrome [34, 35]. In addition, various types of psychiatric problems in ICU survivors may present as sleep disorder, depression, anxiety, or substance abuse disorder. In the current study, we evaluated the number of patients who experienced cognitive impairment or various types of psychiatric problems and documented cognitive impairments that often manifest alongside depression, sleep disorder, or anxiety. Similar to our results, previous studies have reported that depression, anxiety, and post-traumatic stress disorder are associated with worsening cognitive function [36,37,38]. Bruck and colleagues reported a correlation between self-rated cognitive function and psychological symptoms [36]. Mikkelsome and colleagues reported that cognitive dysfunction was often associated with increased anxiety in long-term survivors of acute respiratory distress syndrome who received ICU care [38]. Importantly, it was found that one ICU survivor often experienced various psychiatric problems within 2 years. Therefore, monitoring for detailed cognitive impairment or psychiatric problems is necessary to provide appropriate management for PICS.

Delirium increases medical costs

The in-hospital and post-ICU costs were higher in patients who experienced delirium than in those who did not. In addition, the risk of increased costs related to cognitive impairment or psychiatric problems after discharge was significantly higher in patients with delirium than in those without delirium (HR, 1.50; 95% CI 1.44–1.56). Previous studies have identified a link between delirium and in-hospital cost. Milbrandt and colleagues conducted prospective cohort study of mechanical ventilated patients, and revealed that delirium is associated with significantly higher intensive care unit and hospital costs [39]. Similar results were seen in the post-hoc analysis of the BRAIN-ICU cohort [40]. This study found that cost of delirium increased after adjusting for time-varying severity of illness and length of stay. Increased medical costs may be linked to the characteristics of patients who experienced delirium during hospitalization involving ICU care. These patients may have increased comorbidities, a need for advanced interventions, and increased ICU and hospital lengths of stay. One interesting finding in this study is that delirium affect post-ICU costs. This results might be correlated with increased hospital readmissions and higher post-ICU medical costs [41, 42]. These results suggest that efforts to prevent delirium should be made to reduce the overall ICU-related medical costs.

Limitations

Several limitations of this study need to be acknowledged. First, owing to the use of retrospective national registry data, detailed clinical data, including the reasons for ICU admission and laboratory tests, could not be collected. Second, the HIRA does not provide any survival information after hospital discharge; thus, long-term mortality outcomes could not be evaluated. Third, there might be difference in severity between the two groups, which may have influenced the results. To minimize the risk of severity imbalance, we adjusted mechanical ventilation, ECMO, hemodialysis, and vasopressor drugs. Fourth, since delirium and cognitive impairment were identified based on diagnostic codes with medication, the incidence could underestimation regarding the mild cases. Therefore, delirium was defined as the presence of psychiatric disease codes or a code for antipsychotic or anxiolytic medications. In addition, a sensitivity analysis was performed in the presence of the code for delirium. Fifth, we did not have information on cognitive function at admission, which could be potential confounders for increased cognitive impairment risk after discharge. However, we tried to exclude severe cognitive impairment at admission as including only patient without a history of any psychiatric clinic visit or psychiatric medication. Finally, since data on clinical outcomes was based on claims data, there might be misclassification of cognitive impairment and hard to know the exact date at onset of cognitive impairment. However, the NHIS routinely audits the claims, and the data for clinical outcomes are considered highly reliable and have been used in numerous peer-reviewed publications.

Conclusion

In summary, this study evaluated the impact of delirium on cognitive impairment and psychiatric problems. In patients who received ICU care, those who had delirium during hospitalization had an increased risk of experiencing cognitive impairment or psychiatric problems after discharge. Many patients showed multiple cognitive impairments and psychiatric problems during the follow-up period. The in-hospital and post-ICU costs were higher in patients who experienced delirium than in those who did not.