key: cord-0926918-0ei76hor
authors: Ahlström, Björn; Frithiof, Robert; Hultström, Michael; Larsson, Ing‐Marie; Strandberg, Gunnar; Lipcsey, Miklos
title: The swedish covid‐19 intensive care cohort: Risk factors of ICU admission and ICU mortality
date: 2021-02-08
journal: Acta Anaesthesiol Scand
DOI: 10.1111/aas.13781
sha: 79be8f5adb225cc1b188a0d61e8551eedb01e55e
doc_id: 926918
cord_uid: 0ei76hor

BACKGROUND: Several studies have recently addressed factors associated with severe Coronavirus disease 2019 (COVID‐19); however, some medications and comorbidities have yet to be evaluated in a large matched cohort. We therefore explored the role of relevant comorbidities and medications in relation to the risk of intensive care unit (ICU) admission and mortality. METHODS: All ICU COVID‐19 patients in Sweden until 27 May 2020 were matched to population controls on age and gender to assess the risk of ICU admission. Cases were identified, comorbidities and medications were retrieved from high‐quality registries. Three conditional logistic regression models were used for risk of ICU admission and three Cox proportional hazards models for risk of ICU mortality, one with comorbidities, one with medications and finally with both models combined, respectively. RESULTS: We included 1981 patients and 7924 controls. Hypertension, type 2 diabetes mellitus, chronic renal failure, asthma, obesity, being a solid organ transplant recipient and immunosuppressant medications were independent risk factors of ICU admission and oral anticoagulants were protective. Stroke, asthma, chronic obstructive pulmonary disease and treatment with renin‐angiotensin‐aldosterone inhibitors (RAASi) were independent risk factors of ICU mortality in the pre‐specified primary analyses; treatment with statins was protective. However, after adjusting for the use of continuous renal replacement therapy, RAASi were no longer an independent risk factor. CONCLUSION: In our cohort oral anticoagulants were protective of ICU admission and statins was protective of ICU death. Several comorbidities and ongoing RAASi treatment were independent risk factors of ICU admission and ICU mortality.

Methods: All ICU COVID-19 patients in Sweden until 27 May 2020 were matched to population controls on age and gender to assess the risk of ICU admission. Cases were identified, comorbidities and medications were retrieved from high-quality registries.

Three conditional logistic regression models were used for risk of ICU admission and three Cox proportional hazards models for risk of ICU mortality, one with comorbidities, one with medications and finally with both models combined, respectively.

We included 1981 patients and 7924 controls. Hypertension, type 2 diabetes mellitus, chronic renal failure, asthma, obesity, being a solid organ transplant recipient and immunosuppressant medications were independent risk factors of ICU admission and oral anticoagulants were protective. Stroke, asthma, chronic obstructive pulmonary disease and treatment with renin-angiotensin-aldosterone inhibitors (RAASi) were independent risk factors of ICU mortality in the pre-specified primary analyses; treatment with statins was protective. However, after adjusting for the use of continuous renal replacement therapy, RAASi were no longer an independent risk factor.

In our cohort oral anticoagulants were protective of ICU admission and statins was protective of ICU death. Several comorbidities and ongoing RAASi treatment were independent risk factors of ICU admission and ICU mortality. Some studies have investigated whether renin-angiotensinaldosterone system inhibitors (RAASi) and other antihypertensive agents could predispose individuals to severe COVID- 19 . 1 Yet, data are scarce on the effect (beneficial or harmful) of these drugs on outcome. Moreover, corticosteroids have been suggested to both elevate and reduce the risk of adverse outcomes in COVID-19. 2, 3 Finally, because thromboembolic complications are common in COVID-19 patients, anticoagulant treatment could be protective against severe Because data are limited on the role of pre-existing diseases and medications on the risk of severe illness and death in COVID-19 patients, we conducted a register-based study on all Swedish COVID-19 patients treated in an intensive care unit (ICU). These patients were compared to an age-and sex-matched cohort.

Our primary outcome was the impact of pre-COVID-19 medications and comorbidity risk of ICU admission as a proxy for severe illness in COVID-19. We also assessed the effect of these factors on the risk of ICU mortality. 

The study population was defined by a least one COVID-19 registration in the SIRI until data acquisition on 27 May 2020. From RTB, four age-and sex-matched controls per patient were drawn. Age matching was performed as close to ICU admission as possible, on the age at 31 January 2020. Cases could not become controls and controls could not be selected twice. Exclusion criteria were aged <18 years or the absence of a Swedish personal identification number (PIN).

The NPR provided data on all contacts with specialized care (eg admission date, discharge date, interventions and diagnoses according to the International Codes of Diagnoses, ICD-10) from 5 years preceding the inclusion date. Data on all dispensed drugs (such as the Anatomic Therapeutic Chemical classification system (ATC) code, dose, strength, number of doses and dispensation date from 2 years preceding inclusion) were retrieved from the Swedish Prescribed Drug Register. We received data on intensive care interventions and status at discharge (dead or alive) from the SIR.

For descriptive statistics, we used medians with interquartile range (IQR), counts with percentages. Differences were evaluated with the Mann-Whitney U test and Fisher´s exact test as appropriate. In the case control cohort, we used three conditional binary logistic regression models and in the ICU discharged cohort we used three binary logistic regression models, to determine the odds ratio (OR) of ICU admission and ICU death, respectively. In the first model we assessed the impact of predefined comorbidities and in the second we assessed the effect of medications in which drug dispensation in the past 6 months preceding inclusion was used as a surrogate for drug use (Appendix 1). In the third model we combined the comorbidities model and the medications model. Immunosuppressed disease or state was then exchanged for systemic inflammatory disease and solid organ transplant recipient because immunosuppressant use, including corticosteroids, was part of the definition of an immunosuppressed state. In the medications models we adjusted for the revised Charlson Comorbidity Index (CCI), 9 which was used as a factor. In three corresponding Cox proportional hazards models we adjusted for gender, age and the Simplified Acute Physiology Score 3 (SAPS3).

Age and SAPS3 were treated as continuous variables after restricted cubic spline application. Observations were censored at the date of alive ICU discharge or at the 27 may 2020, whichever occurred first.

The ICD-10 codes for comorbid diagnoses and the ATC codes for medications appear in Appendix 2.

As sensitivity analyses we performed the combined regression 

In the SIRI there were 2786 care episodes with COVID-19 representing 1981 patients > 17 years old with a PIN included in the RTB from which 7924 population controls were drawn. On the date of data extraction, 1544 patients had been discharged from the ICU ( Figure 1 ).

the ICU care of the cohorts are described in Table 1 .

The crude occurrence of all analysed comorbid diseases were more common in the ICU-admitted COVID-19 patients than in the controls, except for stroke and cancer. In addition, alpha-blockers, tumour necrosis factorα (TNFα) inhibitors, interleukin inhibitors, oral anticoagulants, Lopinavir/Ritonavir and anti-hepatitis C virus (HCV) drugs were not associated with ICU admission. All other medications were more commonly dispensed to the ICU-admitted patients than to the controls ( Table 2 ). Amongst the ICU discharged patients, the crude occurrence of ischemic heart disease, hypertension, type 2 diabetes mellitus (T2DM), stroke, chronic obstructive pulmonary disease (COPD) and being immunosuppressed were more common in deceased than in ICU survivors. For the remaining comorbidities, there were no differences. RAASi, beta-blockers, non-insulin antidiabetics, immunosuppressants, statins and platelet aggregation inhibitors were more commonly dispensed to patients who ultimately died in the ICU. No other differences were seen for the remaining medications analysed ( Table 2 ).

In the comorbidities logistic regression model, hypertension, T2DM, chronic renal failure (CRF), asthma, obesity and being immunosuppressed were independent risk factors for ICU admission for COVID-19 ( Figure 2 ).

In the logistic regression model with medications RAASi, statins and immunosuppressant medication, including glucocorticoids, were independent risk factors for ICU admission. In the combined comorbidities and medication model hypertension, T2DM, CRF, asthma, obesity, being a solid organ transplant recipient and immunosuppressants, including glucocorticoid therapy, remained independent risk factors of ICU admission. In the same model anticoagulant therapy was protective of ICU admission ( Figure 3 ). The list of oral anticoagulants used in the cohorts is found in Appendix 3. In these three conditional logistic regression models, Lopinavir/Ritonavir and Anti-HCV and/or Interferon were excluded due to infinite beta.

In the comorbidities Cox model asthma, higher SAPS3 and age were independent risk factors of death, whereas non-ischemic heart disease was protective. In the medications Cox model RAASi therapy, higher SAPS3 and age were independent risk factors and statins as well as oral anticoagulant therapy were protective of ICU mortality.

Because of no cases with anti HCV therapy, this was excluded from the model (Figure 2 ). In the Cox model combining comorbidities and medications increasing SAPS3 and age, stroke, COPD, asthma and RAASi therapy remained independent risk factors of ICU death, and statins remained protective ( Figure 3 ). Due to missing data, 256 patients had SAPS3 and 36 patients had time at risk imputed by MICE.

The crude occurrence of the variables used in our models, divided on the ICU admitted patients, the ICU discharged patients and the

patients not yet discharged from ICU is shown in Table 3 . CCI score 0 (0-0) <.001 0 (0-0) 0 (0-0) 0 (0-0 disease was protective. Stroke was no longer a significant risk factor and Statins were no longer protective (Appendix 7). current study is that although obesity was a risk factor for ICU admission, it was not a risk factor for ICU death. This is in line with the findings from a smaller study assessing obesity as a risk factor of illness severity in COVID-19 patients 18 and also studies on sepsis. 19 Furthermore, it supports the findings from a large Brazilian cohort explored for risk factors of COVID-19 mortality. 20 Although COPD has been suggested as a risk factor for severe COVID-19 disease, 21 this was not confirmed in our main analyses on risk of ICU admission but it was associated with ICU mortality. However, asthma was a risk factor across all analyses. Perhaps severe COPD patients are more often subject to limitations of care or we do not adjust for some protective factors related to COPD.

In contrast with previous studies, RAASi were associated with an increased risk of ICU death after adjusting for comorbidities.

On the assumption that acute renal failure would be a mediator we added use of CRRT to the model. However, RAASi was still, although to a lesser extent, associated with ICU mortality, suggesting that acute renal failure could not explain the association between has been reported being a risk factor of ischemic stroke in previous studies 25 and stroke has been linked to COVID-19 mortality in a general population cohort. 26 Here, we report that previous stroke is a risk factor of COVID-19 related ICU mortality.

One important limitation of the present study is that patients with an indication for ICU care may not be admitted because of capacity strain and care limitations. High age and severe comorbidities are reasons for such limitations, which might skew the results.

However, this constraint is inherent in ICU care. The surge might also have altered usual indication of ICU care with an increased use of high flow oxygen and noninvasive mechanical ventilation on the hospital wards. To address this we performed a sensitivity analysis on patients with IMV during ICU. This conditional logistic regression rendered the same results as our primary analysis on ICU admission apart from oral anticoagulants not being protective. However, for a significant proportion of the ICU-admitted patients we lack information on IMV, which might affect the results. In addition, the control population was not matched for residence, which may affect the prevalence of the tested prognostic factors that vary somewhat between different parts of Sweden. 27 However, these differences may have been related to age differences between regions of Sweden and for our study we used age-matched controls. Lastly, the lack of data on frailty, which is an important risk factor of ICU mortality, especially in the aged, 28 is also an important limitation.

This study has several strengths. We compared the COVID-19 ICU population to a relevant control cohort of the general population to create a robust dataset for analysis of risk factors and protective factors of severe COVID-19 defined by ICU admission and death. The cohort and all data in our dataset are derived from high-quality national registries with a low rate of missing data. We believe that this approach is superior to identifying controls by a negative COVID-19 test, which is a common strategy used by others. 1, 29 We sought to include severe COVID-19 and the low median arterial partial oxygen pressure divided by the fraction of inspired oxygen (PaO2/FiO2) at ICU admission suggests severe hypoxic respiratory failure in the ICU-admitted cohort. Moreover, the high proportion of invasively ventilated patients amongst the dataset is also proof of a high degree of a physiological disorder. Lastly, we had a relatively high degree of missing SAPS3 data. The missingness is due to the fact that 437 patients were still not discharged at date of data collection and that several ICUs report SAPS3 data to the SIR only after ICU discharge of the patient. We chose to impute these data due to the high risk of bias related to the missingness. The stability of the imputation was tested with a sensitivity analysis on complete cases. The results of this model were consistent with the complete cases model, but differed in the effect on outcome for being a solid organ transplant recipient and previous stroke. These variables had few observations and wide 95% CI. Also non-ischemic heart disease and previous use of statins differed between the complete vs imputed models. Common to these variables was that their statistical interpretation was sensitive to small changes in the dataset. Due to this we have a high confidence in our results, especially for the majority of the variables that did not change their effect over the sensitivity analysis.

This nationwide matched cohort study on risk factors of COVID-19

found that ongoing therapy with anticoagulants is protective of COVID-related ICU admission and ongoing therapy with statins were protective of ICU death. In addition, T2DM, CRF, asthma, obesity, being a solid organ recipient and ongoing immunosuppression were independent risk factors for ICU admission. Increasing age, stroke, COPD and asthma were associated with ICU death, as was ongoing treatment with RAASi. Our findings may be useful in policymaking on the protection of risk groups of severe COVID-19 infection and ICU death and warrant further research on anticoagulation and RAASi in COVID-19 disease.

The authors have nothing to disclose.

The data used in this study are available from the SIR, the NPR and the SCB. However, privacy or ethical restrictions apply to the availability of these data, which were used under license for the current study. Thus, these data are not publicly available. 

Risks and Impact of angiotensinconverting enzyme inhibitors or angiotensin-receptor blockers on SARS-CoV-2 infection in adults

The effect of corticosteroid treatment on patients with coronavirus infection: a systematic review and meta-analysis

Dexamethasone in hospitalized patients with covid-19 -preliminary report

Confirmation of the high cumulative incidence of thrombotic complications in critically ill ICU patients with COVID-19: An updated analysis

The strengthening the reporting of observational studies in epidemiology (STROBE)

The Swedish Intensive Care Registry year-report

2020

Registers -Swedish national board of health and welfare

Populations Statistics (RTB)

Updating and validating the charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries

Impact of anticoagulation prior to COVID-19 infection: a propensity score-matched cohort study

Prevention of thrombotic risk in hospitalized patients with COVID-19 and hemostasis monitoring

Potential role of statins in COVID-19

Pharmaco-immunomodulatory therapy in COVID-19

Statin use is associated with lower disease severity in COVID-19 infection

Protective role of statins in COVID 19 patients: importance of pharmacokinetic characteristics rather than intensity of action

Atorvastatin associated with decreased hazard for death in COVID-19 patients admitted to an ICU: a retrospective cohort study

Risk factors associated with mortality among patients with COVID-19 in intensive care units in Lombardy, Italy

Pre-existing traits associated with Covid-19 illness severity

The obesity conundrum in sepsis

Risk factors for hospitalization and mortality due to COVID-19 in Espírito Santo State, Brazil

Chronic obstructive pulmonary disease is associated with severe coronavirus disease 2019 (COVID-19): COPD and COVID-19

Renin-angiotensinaldosterone system inhibitors and risk of covid-19

Sanchis-Gomar F. Diabetes mellitus association with coronavirus disease 2019 (COVID-19) severity and mortality: a pooled analysis

COVID-19 in people with diabetes: understanding the reasons for worse outcomes

COVID-19 related stroke in young individuals

Factors associated with COVID-19-related death using OpenSAFELY

Burden and prevalence of prognostic factors for severe COVID-19 in Sweden

VIP1 study group. The impact of frailty on ICU and 30-day mortality and the level of care in very elderly patients (≥ 80 years)

Association of use of angiotensin-converting enzyme inhibitors and angiotensin II receptor blockers with testing positive for coronavirus disease 2019 (COVID-19)

The swedish covid-19 intensive care cohort: Risk factors of ICU admission and ICU mortality