Decreased serial scores of severe organ failure assessments are associated with survival in mechanically ventilated patients; the prospective Maastricht Intensive Care COVID cohort

Background The majority of patients with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection are admitted to the Intensive Care Unit (ICU) for mechanical ventilation. The role of multi-organ failure during ICU admission as driver for outcome remains to be investigated yet. Design and setting Prospective cohort of mechanically ventilated critically ill with SARS-CoV-2 infection. Participants and methods 94 participants of the MaastrICCht cohort (21% women) had a median length of stay of 16 days (maximum of 77). After division into survivors (n = 59) and non-survivors (n = 35), we analysed 1555 serial SOFA scores using linear mixed-effects models. Results Survivors improved one SOFA score point more per 5 days (95% CI: 4–8) than non-survivors. Adjustment for age, sex, and chronic lung, renal and liver disease, body-mass index, diabetes mellitus, cardiovascular risk factors, and Acute Physiology and Chronic Health Evaluation II score did not change this result. This association was stronger for women than men (P-interaction = 0.043). Conclusions The decrease in SOFA score associated with survival suggests multi-organ failure involvement during mechanical ventilation in patients with SARS-CoV-2. Surviving women appeared to improve faster than surviving men. Serial SOFA scores may unravel an unfavourable trajectory and guide decisions in mechanically ventilated patients with SARS-CoV-2.

Progressive multi-organ disease increases mortality, although it may be heterogeneous over time and vary between and within individual patients. For example, data suggest that women are less severely affected by SARS-CoV-2 infection than men [3,14]. The course of multiorgan disease could, therefore, also potentially differ between men and women. Furthermore, changes in the number and severity of organ systems involved over time may also include valuable prognostic information that may guide clinical decisions for mechanically ventilated patients.
The Sequential Organ Failure Assessment (SOFA) score, widely established to determine multi-organ failure in Intensive Care Unit (ICU) patients, is designed to evaluate changes in organ failure over time [26][27][28]. The SOFA score includes components reflective of the respiratory, coagulation, liver, cardiovascular, renal, and central nervous systems. Whether trends in SOFA score during ICU admission are associated with outcome remains to be established for SARS-CoV-2 infection [29].

Methods
The manuscript was written following the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guideline [30].

Participants
The Maastricht Intensive Care COVID (MaastrICCht) cohort study design has been described more extensively elsewhere [31]. Briefly, this prospective cohort study was conducted in patients admitted to the Intensive Care of the Maastricht University Medical Centre+ (Maastricht UMC+), a tertiary care university teaching hospital in the southern part of the Netherlands. Usually, the Maastricht UMC+ ICU has 27 beds, divided over three subunits to which all types of critically ill patients are admitted. During the COVID-19 pandemic, our ICU was rapidly stepwise upgraded to a maximum of 64 beds, consisting of six subunits with 52 beds for COVID-19 patients and two subunits with 12 beds for non-COVID Intensive Care patients. The study was designed to foster other datasets and registries according to the FAIR data principle in collaboration [31]. The local institutional review board (Medisch Ethische Toetsingscomissie (METC) 2020-1565/ 300523) of the Maastricht UMC+ approved the study, which was performed based on the regulations of Helsinki. During the pandemic, the board of directors of Maastricht UMC+ adopted a policy to inform patients and ask their consent to use the collected data and stored left-over serum samples for COVID-19 research purposes. The study is registered in the Netherlands Trial Register (registration number NL8613).
This study included all participants with respiratory insufficiency requiring mechanical ventilation and at least one PCR positive for SARS-CoV-2 and/or a chest CT scan strongly suggestive for SARS-CoV-2 infection, based on a CORADS-score of 4-5 scored by a radiologist [32]. Participants were followed until primary outcome was reached (i.e. either death in the ICU or discharge from the ICU). After training by qualified research staff and with daily supervision by a senior investigator, medical research interns and PhD candidates not involved in patient care included participants and collected clinical, physiological, and laboratory variables using a predefined study protocol (extensively described elsewhere) [31]. For the present study, participants were included from March the 25 th , the inception of the cohort, until June the 23 rd 2020.

Multi-organ failure variables
Within the MaastrICCht cohort every component of the SOFA score was collected daily in mechanically ventilated patients with a SARS-CoV-2 infection [31]. The SOFA score includes components reflecting the status of coagulation, the liver and the respiratory, cardiovascular, central nervous, and renal organ systems. Each organ system component is scored as one of five categories, ranging from 0 (normal organ function) to 4 (worst organ function). The SOFA score is the sum of the six organ system component scores and thus ranges from 0 to 24. The SOFA score was developed to evaluate multi-organ function daily, which is a major advantage to study the development of multi-organ failure over time [26]. Evidence for SARS-CoV-2 infection, and definition of SOFA score and its components are shown in Table 1.

Outcome variables
The study population was divided into two subgroups, participants who had died during their ICU stay and participants who were discharged from the ICU alive.

Confounders
Comorbidities were proposed as confounder as these can be associated with organ function at baseline and determine patient outcome [33]. For the present study, in addition to age and sex, chronic lung, liver and renal disease, and COVID-19 related comorbidities, such as obesity (body mass index, BMI kg/m 2 ), diabetes mellitus, and presence of cardiovascular risk factors, present on admission, were considered as potential confounders. Furthermore, we considered the admission Acute Physiology and Chronic Health Evaluation II (APACHE II) score as a potential confounder. The APACHE II score is a physiologically based classification system for measuring severity of illness in groups of critically ill patients [34]. APACHE II and SOFA score differ, although both score severity of critical illness. The APACHE II score was primarily developed to rank disease severity between patients over the first 24 h of admission, whereas the SOFA score was developed to monitor disease severity within a patient over time [26].

Statistical analyses
The sample size was determined pragmatically; all participants eligible for the study that had been enrolled in the cohort until June the 23 rd 2020 were included. The data were analysed with R version 3.6.1. The sample characteristics were described using mean and standard deviation (SD), median and interquartile range (IQR), or percentage, as appropriate.
First, the cohort was categorised into ICU-survivors and ICU-nonsurvivors. All participants reached a primary outcome. We computed estimates of group differences in the trajectory of average SOFA scores over time between those discharged alive, and those who had died. Next, we used linear mixed-effects regression with a random intercept and random slope with time to compute differences in average SOFA scores and differences in the slope over time between both groups. Specifically, we used unstructured variance-covariance matrix and an autoregressive correlation structure of the first order for longitudinal measures. To assess non-linear change over time, we added polynomials of time. Using the Akaike Information Criterion, the best fitting model for change over time was selected.
We computed the crude group differences (Model 1). Next, the model was adjusted for age and sex (Model 2), and additionally for COVID-19 related comorbidities such as obesity (BMI), diabetes  mellitus, and the presence of cardiovascular risk factors and chronic lung disease, chronic liver disease, and chronic renal disease at baseline (Model 3). Subsequently, model 3 was adjusted for the APACHE II score (Model 4), to further disentangle between patient disease severity (APACHE II) from within patient disease severity over time (SOFA score) in the association between disease severity and outcome. We also tested for effect-modification of the association between SOFA score over time and outcome by sex by adding a three-way interaction term to Model 2.
As twelve participants were transferred from ICU because of logistical reasons, we conducted a sensitivity analyses and repeated the main analyses without those 12. We checked the percentage of missing values for all potential confounding variables as determined in the previously published protocol [31]. Data would be imputed if the proportion of incomplete patients is over 5%, excluding the longitudinal measures as they were analysed using generalised linear mixedeffects regression. In case of over 5% of incomplete records, multiple imputation would be performed.

Results
The MaastrICCht cohort includes a total of 94 participants at the time of data extraction. The mean age was 64.3 ± 11.9 years, 21% were women. In total, 1555 serial SOFA scores had been recorded, with a mean of 7.7 ± 2.3 on admission. The mean APACHE II score on admission was 15.8 ± 5.7. Correlation coefficient between SOFA score and APACHE II score on admission was 0.54. All participants reached primary outcome, one of which did not contribute any SOFA score. Of the 93 participants included in the analyses, 35 (38%) had died and 58 were discharged alive (supplemental Fig. S1). The median duration of stay in the ICU was 16 days (1st and 3rd quartile: 8 and 24 days), with a maximum of 77 days. Table 2 shows the characteristics of the included participants stratified by primary outcome. Of all confounding variables, only BMI was missing in 1 (1%) participant. Hence, no data imputation was performed.
The mean SOFA score at baseline was 7.3 for eventual ICU-survivors (SD: 1.9) compared to 8.6 (SD: 2.8) for those who eventually died in  the ICU (p-value for difference: 0.068). Fig. 1 shows the individual trajectories of observed SOFA scores for ICU-survivors and ICU-nonsurvivors. Fig. 2 Fig. 1). Fig. 3 shows development of categories of SOFA scores over weeks per participant for those who survived the ICU, and those who did not survive. Observed and predicted SOFA scores over time for ICU-survivors and in ICU-nonsurvivors. A lower SOFA score indicates less organ dysfunction. Note that some jitter has been applied to the raw data to make multiple observations with the same values visible.

Table 3
Results of linear mixed-effects models: difference in SOFA score development between ICU-survivors and ICU-non-survivors.

Model
Regression coefficient (95% CI) p-value  Adjustment for sex and age, and additionally for the presence of chronic lung disease, chronic liver disease and chronic renal disease, and obesity, diabetes mellitus, and cardiovascular risk factors, did not materially change the results (Table 3; Models 2 and 3). Additional adjustment for the APACHE II score reduced the negative regression coefficient that indicated the overall lower SOFA score over time (−0.99, 95% CI: −1.93; −0.06) for ICU-survivors compared to ICU-non-survivors, but the association remained significant. The improvement in SOFA score over time (−0.18, 95% CI: −0.25; −0.12) did not materially change (Table 3, Model 4).
We observed a significant interaction between sex and the association between SOFA score over time and ICU mortality (p = 0.043). After adjustment for age, compared to nonsurvivors, women survivors had a lower overall SOFA score during their ICU stay (−1.76, 95% CI:  Table 4). Table 5 shows the organ component scores. The PaO 2/ FiO 2 ratio was not associated with outcome. However, after adjustment for age and sex, ICU-survivors showed an overall higher PaO 2 over time (0.78 kPa, 95% CI: 0.19; 1.38), and both an overall lower FiO 2 need (−9.2%, 95% CI: −14.1; −4.3) and a lower FiO 2 need slope over time (−0.59%, 95% CI: −0.97; −0.21), as compared to ICU-non-survivors. After adjustment for age and sex, the SOFA cardiovascular component score did not differ between groups over time (−0.19 points, 95% CI: −0.70; 0.32), but the slope over time for ICU-survivors was lower compared to ICU-nonsurvivors (−0.09 points, 95% CI: −0.15; −0.03). After adjustment for age and sex, ICU-survivors showed both an overall lower SOFA renal component score (−0.83 points, 95% CI: −1.40; −0.26) and a lower SOFA renal component score slope over time (−0.05 points, 95% CI: −0.08; −0.02), as compared to ICU-non-survivors. Bilirubin, the Glasgow coma score and thrombocytes count, indicators for respectively, the liver, the central nervous system, and coagulation components, showed no association with survival.
Sensitivity analyses excluding 12 patients transferred for logistical reasons did not alter any of the conclusions.

Discussion
In this prospective cohort study including 93 mechanically ventilated participants with SARS-CoV-2 infection, we made five main observations. First, a decrease in SOFA score over time (which indicates improved organ function) is associated with ICU survival. Second, the association of the decrease in SOFA score with ICU survival remained present after adjustment for age, sex, the presence of chronic lung, renal and liver disease, obesity, diabetes mellitus, and cardiovascular risk factors, and after adjustment for the APACHE II score. Third, concerning the individual components of the SOFA score; the respiratory, circulatory, and renal organ components [35] appeared the most important drivers of the difference in trajectories of the SOFA score over time between ICU-survivors and ICU-non-survivors. The liver, the central nervous system, and coagulation components did not seem to play a role [36,37]. Fourth, the decrease in SOFA score over time between patients who survived the ICU vs. those that did not was statistically significantly greater for women (steeper slope) than men. Fifth, a higher admission SOFA score was not associated with ICU death.
The APACHE II score was primarily developed to rank disease severity between patients over the first 24 h of admission, whereas the SOFA score was developed to measure changes in disease severity ICU-non-survivors as the reference category. The sample size for each of the analyses is reduced to 73 for the analyses for men and 20 for women. a A negative regression coefficient indicates that the average SOFA score of survivors is overall lower over time compared to the non-survivors. b A negative regression coefficient for the interaction term indicates that the average SOFA score of survivors decreases more over time compared to the non-survivors. i.e. the interaction between group and time models the change over time for both groups separately). over time [40]. The results show that in particular the difference in trajectories of SOFA score over time between ICU-survivors and ICU-non-survivors was independent of APACHE II score on admission. Furthermore, the SOFA score and APACHE II score on admission had a moderate correlation. Adjusting the association between SOFA score and outcome for APACHE II score appears odd as both scores identify disease severity of critical illness. This analysis, however, illustrates the fact that appropriate use of disease severity scores measuring alternative sources of variation (between patients vs. within patients) in multi-morbidity (i.e. both chronic multi-morbidity and acute multiorgan failure) is of utmost importance. The observation that trajectories of SOFA score are associated with outcome, independent of APACHE II score, could thus help to further refine the recent rapid guideline advice against the use of the SOFA score for ICU triage for patients with COVID-19 [41]. The present results, for example, add that appropriate SOFA score application in critical care, aids to identify patients with a favourable disease course [26,42,43]. This cohort study design has several strengths. First, the study is prospective by design and allows for many serial measurements over time in patients with SARS-CoV-2 infection. Second, systematic data collection is performed using a predefined protocol. Third, sensitivity analyses did not alter conclusions. A limitation of the study is the single centre approach and a relatively small sample size. However, the fast spread of the SARS-CoV-2 virus affects patients world-wide and urgently requires data to guide clinical decisions. Observations made in the MaastrICCht participants with SARS-CoV-2 may be generalised to other critically ill patients only.
Nevertheless, including a heterogeneous sample of patients admitted to the ICU, without further exclusion criteria, reduced the chance of selection bias and contributes to the internal validity of the results for mechanically ventilated patients with SARS-CoV-2 infection. The SOFA score components use only a limited set of variables per organ component, and as weighting is applied to each component score, the overall SOFA score likely reflects pathophysiology of true multi-organ dysfunction suboptimally. Using a limited set of variables could have led to an underestimation of the reported association between multiorgan failure over the course of time and survival. Although multiple more sophisticated risk scores will be developed using traditional and artificial intelligence techniques, [44] SOFA score is widely known and easily applicable at the bedside. The latter features of the SOFA score are essential when resources are scarce, and time is of the essence in crises like the COVID-19 pandemic.
In summary: The outcome of patients with a SARS-CoV-2 infection admitted to the ICU is unfavourable for many. Admission characteristics seem insufficient to guide decisions about whether or not patients are likely to survive. This study revealed that temporal changes in multiorgan systems yield information that may guide decisions in individual mechanically ventilated patients with a SARS-CoV-2 infection [44]. The temporal change in SOFA score could be considered contributory to a decision to continue life-sustaining treatment or forgo life-sustaining support if considered futile. Caregivers can initiate adequate and timely end-of-life care and support. Furthermore, optimisation of care could have beneficial effects on caregivers and even the availability of beds for new patients in need of care. The extent of decrease in SOFA score during ICU admission that enables to predict outcome in mechanically ventilated patients with SARS-CoV-2 warrants further study in larger datasets [45].

Conclusions
Multi-organ involvement is a predominant characteristic of the SARS-CoV-2 infectious disease course in mechanically ventilated patients. A decrease in SOFA score over time, indicating improved organ function, is associated with ICU survival. The association between decreased SOFA score over time and survival was independent of comorbidities. Concerning the individual components of the SOFA score; the respiratory, circulatory, and renal organ components appeared most important. The results were more pronounced for women than men. Admission SOFA score was not associated with ICU death. These results suggest that SARS-CoV-2 infection can include multi-organ dysfunction that has a heterogeneous course with many dimensions. Serial SOFA scores may help guide optimisation of individual patients' critical care in case of a second wave of the COVID-19 pandemic.

Ethics approval and consent to participate
The local institutional review board (Medisch Ethische Toetsingscomissie (METC) 2020-1565/ 300,523) of the Maastricht UMC+ approved the study, which was performed based on the regulations of Helsinki. During the pandemic, the board of directors of Maastricht UMC+ adopted a policy to inform patients and ask their consent to use the collected data and stored left-over serum samples for COVID-19 research purposes. The study is registered in the Netherlands Trial Register (registration number NL8613).

Consent for publication
Not applicable.

Availability of data and material
No concrete agreements on data sharing have been made yet. Before any data is shared outside the MUMC+, a datasharing plan will be drawn up in consultation with the data officer that conforms to relevant laws and regulations concerning personal data.

Funding
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Authors' contributions
IH and BB conceived and designed the study. JB, JT, RG, CG, RS, and SM contributed to data collection. SK, JB and BB analysed the data. JB, SK, CG, WM, IH, and BB drafted the manuscript. FT, JT, RG, RS, MA, MP, SM, and DB critically reviewed the manuscript. All authors read and approved the final manuscript.

Declaration of Competing Interest
The authors declare that they have no competing interests.