This cohort study of over 1.8 million people used the Region Stockholm real-time Covid-19 monitoring framework, allowing for detailed analysis of risk factors for Covid-19 related mortality, hospitalization and ICU-admission. The key findings include that across age groups and irrespective of sex, kidney failure, diabetes type II and obesity increase risk of mortality. In contrast, cardiovascular comorbidities diverge; heart failure and ischemic heart disease are risk factors for death, but atrial fibrillation and hypertension are not. Risks of hospitalization with regards to comorbid conditions follow a similar pattern, whereas admission to intensive care differs. Triage flow processes were clearly present as demonstrated by hazard ratios for heart failure, IHD and kidney failure; strongly associated with death and hospitalization but not with ICU admission. Ages 80 years and up accounted for two thirds of all Covid-19-related deaths in Stockholm.
Initial studies of risk factors have focused solely on patients with Covid-19. In hospitalized patients from Wuhan increasing age, higher illness severity score and increased d-dimer were associated with risk of in-hospital death (18). A systematic review detailed impact of comorbidities on the disease course of Covid-19. It included 61 cohort studies and 31,089 (median 162; IQR: 103–338) patients were included in the meta-analysis. Cerebrovascular disease, chronic obstructive pulmonary disease, cardiovascular disease, hypertension, diabetes mellitus and malignancy were risk factors for poor clinical outcome in Covid-19 (5). As mentioned in the introduction, the two definitions of “poor clinical outcome” of Covid-19 across these studies were either severity of Covid-19 or admission to intensive care. Both outcomes lack rigid definitions. ICU admission is prone to regional or local variability as it is affected by medical tradition as well as ICU bed availability. Only 12 of these 61 investigations report mortality.
The present investigation has an advantage: insight into the co-morbid burden for a study population comprising over 1.8 million individuals (see Tables 1–3). Similarly, the UK OpenSAFELY study had access to primary care records of over 17 million adults, linked to 10,926 Covid-19-related hospital deaths (3). Increasing age, deprivation, being male; as well as diabetes, severe asthma, liver disease, kidney disease and multiple other comorbidities were associated to increased risk of dying of or with Covid-19. Using data from the real-time Covid-19 monitoring framework in Region Stockholm we have built a comprehensive overview on diagnosis from primary care, inpatient, and outpatient specialist care as well as information on conformed Covid-19 from a large, well-defined health care region. The present study differs from the UK study with regards to some key elements: firstly, we have complete coverage of all adult Stockholm residents. Secondly, we have data on both hospital- and out-of-hospital mortality and not only hospital mortality. Lastly, we report risk of hospitalization and ICU admission. This tertiary outcome, admission to intensive care, missing from the UK study, must however be analyzed with caution. Investigating co-morbid conditions as risk factors and their association to a soft outcome like ICU admission is always tricky, and more so during a pandemic. ICU admission is affected by ICU capacity. In a study from 2012, Sweden had 5.8 ICU beds/100 000 inhabitants as compared with the EU average of 11.5 (19). Notably, the Stockholm ICU bed density is even lower at 4.2/100 000 (20).
Despite the fact that emergency changes to increase intensive care capacity were implemented (21) it was paramount to select patients with the highest chance of benefitting from ICU admission.
The results of the real-time Stockholm Covid-19 monitoring framework, combined with rigorous modeling, adjusting for age, sex and socioeconomic status using causal diagrams, show how comorbid conditions are or are not associated with risk of hospitalization or death. This study relies on the accuracy of diagnoses reported in medical records in which some might be missing or misclassified. However, the diagnostic validity of recorded diagnoses in Sweden in is generally high (12). This allows for development of predictive models that in turn can be used for granular health care planning. Naturally, despite efforts to control for associations with certain co-morbidities and SARS-CoV-2 exposure and measures taken to minimize risk of bias, causality cannot be proven. Clearly, multiple observations in the present investigation indicate the need for rigorous and focused studies; are patients with atrial fibrillation protected by the fact that they almost always medicate with warfarin or novel oral anticoagulants? This question deserves further attention as several reports show that thrombotic events in hospitalized Covid-19 patients are common and dangerous (22). Indeed, the in-hospital/in-ICU clinical practice has undergone rapid changes, with increased use of anticoagulation. Our study sheds more – but not enough – light on hypertension; a non-significant risk factor for most age strata in our adjusted models. Interestingly, hypertension has been reported to have an inverse association with mortality among elderly UK patients (3) but was very frequent in both US and Chinese case series (23). Focus could also be devoted to how kidney failure seemingly affects ICU patient selection in males and females.