Risk factors for readmission of COVID-19 ICU survivors: A three-year follow up

Materials & Methods

Results

We analysed data from 505 consecutive study participants admitted to the ICU from March 4, 2020, to March 22, 2021, with severe COVID-19. Of these, 164 (32.5%) died in the ICU. The demographic characteristics and comorbidities of the ICU survivors have been shown in table I. The median age was 60 yr (IQR: 50–68), with a predominance of males (68.9%) and most having a body mass index (BMI) above 25 Kg/m² (84.8%). The common comorbidities were hypertension (44.6%) and dyslipidaemia (43.7%).

The median time from symptom onset to hospital admission was six days (IQR: 4–8), and from hospital to ICU admission was three days (IQR: 1-5). The most common symptom in the emergency department was fever (85%). Most of these study participants (82.1%) exhibited a bilateral infiltrate on the first chest X-ray performed upon hospital admission. The median NSS on admission was 15 (IR: 11-18).

Table II provides a summary of the treatment course and outcomes observed in the ICU. The median ICU length of stay was 14 days (IQR: 8-33.5). During this period, nearly 60 per cent of study participants required mechanical ventilation, with a median duration of 20 days (IQR: 11–44).

Among the study participants who survived their ICU stay, 22 per cent (n=75) required readmission following their initial hospital discharge. The median follow up period was 1,154 days (IQR: 1,054–1,252 days), equating to more than three years. Figure presents a detailed flowchart illustrating study participants readmission.

Close

Figure.

Flowchart of study participants’ readmission. ICU, intensive care unit.

Export to PPT

The median time to hospital readmission was 415 days (IQR: 166–797). Most of the readmissions, 54 study participants (72%), were classified as urgent. According to the MTS performed in the emergency department, more than 50 per cent of these study participants were assigned third-level priority (urgent, yellow), while 22.2 per cent were classified as having higher clinical risk (very urgent or immediate priority). The full classification is presented in supplementary table I. The most common presenting complaint, according to the MTS flowchart, was ‘unwell adult’ (Supplementary Table II).

Table III outlines the causes of hospital readmission, with respiratory conditions being the most frequent (29.3%). The median length of stay during readmission was six days (IQR: 3-11). The median NSS at the time of readmission was 19 (IQR: 16-20). Only 13 study participants (17.3%) developed new comorbidities not documented during their initial admission (described in Supplementary Table III).

Table IV compares the characteristics of study participants who were readmitted to the hospital with those who were not. Study participants requiring readmission were older and had a higher prevalence of comorbidities, including diabetes, dyslipidaemia, chronic obstructive pulmonary disease (COPD), and a history of cancer. During ICU admission, these study participants exhibited higher creatinine levels, lower PaO₂/FiO₂ ratios, and lower troponin-I levels. Additionally, they more frequently required mechanical ventilation and had longer stays in both the ICU and the hospital.

Multivariate analysis identified three factors independently associated with hospital readmission: advanced age, elevated creatinine levels (greater than 1.2 mg/dl) at ICU admission, and prolonged ICU stay. These findings have been summarized in supplementary table IV.

Of the 75 study participants readmitted to the hospital, 19 required subsequent ICU readmission. The causes for ICU readmission have been detailed in table III, with respiratory issues being the predominant reason (63.2%).

Ten study participants died following their initial hospital discharge, nine of whom passed away during their hospital readmission. The median time from discharge to death was 794 days (IQR: 386–965). Respiratory failure was identified as the cause of death in four study participants, two of whom had residual fibrosis from a prior COVID-19 infection, which was considered an intermediate cause of death. Three study participants died from complications associated with stage IV oncologic disease, one from haemorrhagic shock secondary to ruptured oesophageal varices, and one during aortic valve replacement surgery. For one study participants, the cause of death could not be determined.

Discussion

In this prospective study, we observed a high hospital readmission rate (22%) among survivors of severe COVID-19 infection over a three-year follow up period. Readmission was independently associated with advanced age, elevated creatinine levels at ICU admission, and prolonged ICU length of stay. The predominant cause of urgent readmissions was respiratory, likely reflecting impaired pulmonary function, as most patients did not have pre-existing respiratory co-morbidities. However, long-term ICU readmission and mortality rates were relatively low (5.6% and 2.3%, respectively), with only a few deaths directly attributable to COVID-19. These findings, in conjunction with other reports on the long-term sequelae of COVID-19^2,5, underscore the chronic nature of a disease that is highly lethal in its acute phase.

Patients who recover from COVID-19, particularly ICU survivors, face significant long-term impacts on their quality of life due to physical, psychological, and social sequelae, which increase their risk of hospital readmission. Consequently, understanding the long-term consequences of COVID-19 and identifying risk factors for unplanned hospital readmissions are essential, particularly as the number of survivors continues to grow. While considerable research has focused on readmission risk prediction tools for COVID-19 patients, their efficacy remains uncertain, and few studies specifically target ICU survivors^7,8,15,16.

The NSS evaluates the following parameters on a scale from 1 to 5: physical condition, mental condition, activity, mobility, and incontinence. Although initially conceived as a pressure ulcer risk prediction scale, it has long been established as a prognostic scale for hospitalised patients, with its score being associated with the risk of complications and mortality^11,17. In our study, readmitted study participants had a median NSS of 19 (IQR: 16-20), compared to a median of 15 (IQR: 11-18) during their initial admission, indicating that the severity of readmissions was lower. Nevertheless, a significant proportion of these study participants required ICU readmission.

Our observed readmission rate (22%) exceeds those reported in other Spanish and international studies, which average around 10 per cent^18-20. However, these studies focused on less than one month follow up periods, and readmission rates rise markedly with longer follow up durations^7,8. In international studies with follow up periods exceeding one month, such as those by Günster et al²¹ and Donnelly et al²² readmission rates similarly exceeded 20 per cent, consistent with our findings. Notably, none of these studies specifically examined ICU survivors, a population with a higher prevalence of sequelae^3,5. Our extended follow up period of over three years revealed a median time to readmission of 415 days, emphasising the importance of conducting studies beyond one year from ICU discharge.

The risk factors for readmission identified in our study were age, elevated creatinine levels, and prolonged ICU stay. These factors were aligned with the findings from studies on hospital survivors over shorter follow up periods. Age, a critical determinant of COVID-19 severity and mortality^1,23-25, is also associated with slower recovery of functionality after hospitalisation and a reduced capacity to adapt to disease sequelae. Previous studies have identified age as a predisposing factor for symptom chronification^15,26. For instance, Mooney et al²⁷ observed a nearly two-fold increase in readmission rates among older patients, while Günster et al²¹ and Verna et al²⁸ reported higher risks among individuals over 80 and 60 yr, respectively.

Renal involvement is frequently observed in COVID-19, with acute kidney injury (AKI) occurring in an estimated 3–5 per cent of hospitalised patients^29,30. Among critically ill patients, up to 30 per cent experience grade II–III acute renal failure according to AKI classifications^1,25,31. AKI is associated with disease severity, increased risk of ICU admission, and mortality³². Verna et al²⁸ identified AKI at hospital admission as a risk factor for rehospitalisation among COVID-19 survivors. Similarly, Yeo et al³³ and Nematshahi et al ³⁴ reported that peak serum creatinine levels ≥1.29 mg/dl and 1.2 mg/dl, respectively, during hospitalisation, more than doubled the risk of readmission.

The reason for the association between increased creatinine levels and readmission remains unclear. We consider that renal dysfunction may be both a consequence and marker of a proinflammatory systemic state. Patients with acute renal failure likely present more extensive disease associated with inflammation and immunosuppression and, therefore, present multiorgan complications and a prolonged recovery, incrementing their risk of readmission^28,33,34.

ICU length of stay has not been widely evaluated as a predisposing factor for readmission, as most studies have not specifically focused on ICU populations. The relationship between hospital length of stay and readmission remains contentious. While studies with follow up periods shorter than one month, such as that by Parra et al¹⁸ associated shorter hospital stays with higher readmission risk, in studies with a longer follow up, the risk was inverted³⁵. One potential explanation is that short-term readmissions often result from acute complications of the disease that were not identified at the time of the initial hospital discharge. In contrast, long-term readmissions, not captured in studies with limited follow up periods, could encompass hospitalisations arising from the sequelae and prolonged consequences of the infection. Symptoms of post-COVID syndrome, which can persist beyond one year after infection, likely contribute to long-term hospitalisation needs.

Prior studies identified comorbidities such as hypertension, diabetes, chronic kidney disease, and immunosuppression as potential risk factors for readmission^7,8,28. In our study, the group of readmitted study participants presented more frequently with arterial hypertension, diabetes, immunosuppression, dyslipidaemia, COPD, and a Charlson Index greater than 0. However, these factors did not attain statistical significance in our logistic regression analysis, likely due to the limited sample size. Further investigations with larger cohorts of ICU survivors are necessary to validate these findings.

Despite the strengths of this study, various limitations must be acknowledged, which may impact the interpretation and generalisability of the findings. Firstly, this study was conducted at a single centre in Spain, which might have specific diagnostic and therapeutic practices. Previous research highlighted that healthcare system characteristics could influence readmission rates^7,20,21. The findings from a public hospital within the Spanish healthcare system might differ from those observed in countries where private healthcare systems predominate.

Secondly, the study participants included in this study were treated in 2020 and 2021, when therapeutic approaches to COVID-19 differed from those subsequently recommended. Moreover, vaccination was not yet available during the data collection period. While these factors may limit the extrapolation of these findings to the current context, our data suggest that the treatment regimens employed at the time had minimal impact on post-COVID readmissions. Nonetheless, these results remain valuable for understanding outcomes in COVID-19 survivors admitted to the ICU before 2021.

Thirdly, data on readmissions were extracted from medical records within the Community of Madrid. Admissions to hospitals in other regions might have been missed. However, we consider such cases to be improbable and unlikely to affect the conclusions significantly.

Lastly, distinguishing between individuals hospitalised due to the long-term consequences of COVID-19 and those with COVID-19 who were admitted for unrelated reasons is inherently challenging. Despite this, the observed readmission rate far exceeds both the expected risk in the general population and the historical risk associated with other respiratory infections¹⁵, supporting a causal relationship between COVID-19 and hospital readmission.

In conclusion, patients requiring ICU admission for COVID-19 are frequently readmitted after their initial hospital discharge. Future research should include experimental studies aimed at mitigating the dominant risk factors and extending data collection to multi-centre populations to enhance the generalisability of findings. Our study highlighted the substantial long-term burden faced by survivors of severe COVID-19, particularly those requiring ICU admission. The observed hospital readmission rate underscored the critical need for continued post-discharge care and monitoring, especially for patients with advanced age, renal dysfunction, and prolonged ICU stays, which were identified as independent risk factors. These findings emphasise the importance of tailoring follow up strategies to mitigate the chronic impact of COVID-19 and improve long-term outcomes. Further research is warranted to explore interventions that could reduce readmissions and enhance recovery for this vulnerable population.