Since December 2019, a pneumonia caused by a novel coronavirus (SARS-CoV-2) has emerged around the globe, leading to important health concerns worldwide [1
The most prevalent comorbidities in the Corona Virus Disease 2019 (COVID-19) patients have been reported to be hypertension, diabetes, cardiovascular diseases, and respiratory system disorders [2
]. Between 11 December 2019 and 31 January 2020, of the 1590 cases analyzed by Guan et al. [3
] 25.1% of laboratory-confirmed individuals had at least one comorbidity, with HR of 1.79 (95% CI 1.16–2.77) among patients with at least one comorbidity and 2.59 (95% CI 1.61–4.17) among patients with two or more comorbidities compared with patients without comorbidity. From the 3912 deaths that occurred in England and Wales relating to COVID-19 (March 2020), 3563 (91%) had at least one pre-existing condition [4
Outcomes such as admission to an intensive care unit (ICU), invasive ventilation, death and greater disease severity have been studied in the context of COVID-19 disease [5
]. The findings of these studies suggested that the majority of patients who had moderate to severe respiratory failure required invasive mechanical ventilation and patients with severe illness required admission to the ICU. Furthermore, patients with comorbidity are at higher risks for poor outcomes: respiratory failure, cardiovascular diseases, diabetes, and kidney injury seem to be highly associated with the death of patients with COVID-19 [3
]. Adverse outcomes were also found in older individuals (≥60 years) [4
] and in males [9
] in data from both China and European countries. It is important to consider that most studies have been performed in China, which may have different demographic characteristics and prevalence of comorbidities compared to the European population, and therefore a different effect and impact of risk factors [9
]. Data describing underlying health conditions among Portuguese patients with COVID-19 have not yet been reported. Limited information has been accessible to describe the presence of comorbidities and outcomes in these individuals.
Globally, there has been an effort to implement several control measures to prevent, prepare, and develop integrated policies against COVID-19. The knowledge of the impact of comorbidities on possible outcomes will allow a more efficient clinical triage of patients by identifying those at greater risk of progression of COVID-19 and more severe complications including death. Furthermore, health services can predict more accurately the need for treatments, facilities, equipment, and necessary staff, according to the risk of patients with a positive test. Considering no vaccine or antiviral treatment for COVID-19 proved effective, public health authorities should prioritize measures and activate specific mechanisms for high-risk patients and strategic community-based networks to protect vulnerable groups of people before and after diagnostic of COVID-19.
In the present study, we performed a comprehensive evaluation of a database with 20,293 COVID-19-infected individuals. The aim of this study was to identify factors and the role of preconditions associated with COVID-19 deaths in Portugal from 1 January to 21 April 2020.
2. Materials and Methods
2.1. Study Design
This study consisted of the analysis of an official public dataset provided by the General Health Directorate of Portugal (DGS) on 27 April 2020. The dataset is available from the General-Directorate of Health in Portugal (DGS), by request and submission of a detailed analysis project and an ethical committee authorization to perform research on the dataset [10
]. Provision of the dataset is then at the discretion of the General Directorate of Health. The data were collected by the SINAVE (National System for Epidemiological Surveillance), which is the national database for mandatory diseases and public health problems notifications, managed by DGS, that provided the data fully anonymized. The research team filled the publicly available form and submitted an analysis project. The project was approved by an Ethics Committee (number 165/2020).
2.2. Data Collected
The provided data concern information between 1 January and 21 April 2020 on infected cases and deaths by COVID-19. A confirmed case was based on a positive polymerase chain reaction test. The primary outcome measure was mortality by COVID-19, measured (registered and confirmed) by Medical Doctors serving as health delegates on the daily death registry. The dataset received on 27 April, encompasses confirmed deaths up to 21 April. Potential deaths from cases registered later in the database (in particular between early April and 27 April) are not confirmed and accounted for in the dataset.
As referenced above, the data were registered in SINAVE, an electronic system for epidemiological surveillance of mandatory diseases that includes laboratory notification. SARS-CoV-2 cases are registered by doctors in SINAVE allowing the electronic communication with local, regional, and national health authorities [11
]. The SARS-CoV-2 case definition used is based on the World Health Organization case definition.
The SINAVE allows epidemiological surveillance and epidemiological data registration including laboratory notification, ongoing treatment, hospitalization, existing health preconditions (morbidity), and death registration.
Individualized health precondition indicator variables were constructed from the two variables provided in the database (“precondition” and “other precondition”).
2.3. General Characteristics and Outcome
The data retrieved include individuals’ demographic characteristics (age, sex, region), COVID-19 disease information (death, recovery, still in treatment, hospitalization, intensive care, respiratory support), and preconditions (Asthma, Cancer, Cardiac disease, Hematological disorder, Diabetes, HIV and other immune deficiency, Kidney disorder, Liver disorder, Neuromuscular disorder, Other precondition and None (absence of precondition)). The data on intensive care and respiratory support include a number of individuals with “unknown” status that we considered to not have had these interventions if they were never hospitalized. This work considers as primary outcome mortality among those with a positive molecular test for COVID-19 on record.
2.4. Statistical Analysis
Descriptive statistics, such as absolute and relative frequencies, mean, standard deviations, and medians were used to summarize univariate characteristics.
Bivariable analyses were performed between the outcome variable (death) and variables potentially associated using the X2 test or the Fisher exact test. Further bivariate analysis was performed to evaluate the difference in distribution between preconditions and sex, hospitalization, and intensive care.
We analyzed regional differences based on the second level of the Nomenclature of Territorial Units for Statistical purposes (NUTS II) classification for Portugal.
Logistic regression models were performed using a generalized linear model with binomial error distribution and logit link function to estimate the crude and adjusted Odds Ratio (OR) with 95% confidence intervals (95% CI) for each potential risk factor. The model performance was assessed using the area under the curve (AUC) and corresponding 95% confidence intervals (95% CI). The significance level was set at 5%. Statistical analysis was performed using R software version 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria).
Our study evaluating a national database of more than 20,000 infected individuals constitutes one of the largest population studies on COVID-19 to date. A total of 14.7% of patients needed hospitalization, associated with a case fatality rate (CFR) of 11.1%, while 1.3% were admitted to the intensive care unit (ICU) with an associated CFR of 10.3%. These figures were somewhat lower when compared to previous systematic reviews, where a 20.3–29.3% and 6.8–13.7% of ICU admission and CFR were reported, respectively [12
]. This difference may be related with a different stage of the epidemic in Portugal at the time of data analysis, and differences in the timing of non-pharmaceutical actions (schools and medical faculties closures and lockdown declared one and two weeks after the first COVID-19 case, respectively) [14
], as the majority of studies included in the systematic reviews were from China. Moreover, it is important to consider the variations of proportions that were found in other individual studies.
This study represents one of the first attempts to understand the lethality of COVID-19 in Portugal from infected cases, proposing risk factors based on multivariable analysis. The risk factors for lethality by COVID-19 were sex (male), advanced age, Kidney disorder, Cardiac disease and Neuromuscular disorder, while the absence of a precondition was associated with a reduced chance of mortality after adjusting for other variables of interest. The principal risk factors for lethality by COVID-19 reported in the present study are supported by the literature. The majority of studies report male patients, older patients, and patients with preconditions at an increased risk of infection and mortality irrespective of the region of the globe [12
]. A recent meta-analysis of thirteen studies registered male individuals (OR = 1.76), age over 65 years old (OR = 6.06), smoking habits (OR = 2.51), and preconditions including Cardiovascular disease (OR = 5.19), Diabetes (OR = 3.68), Respiratory disease (OR = 3.68), and hypertension (OR = 2.72) as significantly higher in critical/mortal patients compared to non-critical patients [15
]. Men registered a 48% increased risk of death by COVID-19 in our study, a result previously observed [9
]. We registered an increased prevalence of preconditions in men when compared to women (all preconditions except Asthma); while the absence of preconditions was 30% more likely in female individuals. Moreover, the fact that women could have stronger responses than men in many infectious pathogens [16
], the likelihood of women to search for health care services more than men [17
], and in the adoption of hygiene practices [18
] may further complement the explanation of this result.
Age was an important predictor for mortality in our study, with lethality-adjusted ORs increasing after 55 years of age. Moreover, there was a thirty-year difference in the average age between fatal (average age of 81 years) and non-fatal cases (average age of 51 years). Age was a predictor for mortality in patients both with and without preconditions, a result supported by a previous multicenter cohort study of 191 individuals where increased odds of in-hospital death were associated with older age (odds ratio 1.10, 95% CI 1.03–1.17, per year increase [19
]). These results are partially explained by the increased burden of pre-conditions in older age groups, where dramatic increases for the prevalence of preconditions (particularly for Cardiac disease, Kidney disorder, and Neuromuscular disorder) were registered in older age groups when compared to individuals of less than 55 years of age. Nevertheless, age remained a risk factor for death in our study even in patients without preconditions, as would be expected given that advanced age is a risk factor for death even in the absence of COVID-19 infection. Indeed, it is possible that the increase in risk of death is more due to aging than to COVID-19 infection proper. Still, the link between older individuals and the likelihood to develop severe and critical cases of COVID-19 has been made before, either due to immunosenescence, malnutrition, or ignoring more easily the early symptoms and consequently missing the best time to seek medical advice [20
]. Furthermore, several studies registered an important association between increased age and COVID-19 severity/fatality [22
]. The vulnerability of elderly individuals is illustrated in recent studies. A retrospective observational study investigating mortality in hospitalized patients with COVID-19 registered the great vulnerability of patients residing in retirement homes, with older age independently associated with mortality when adjusted for other variables of interest [22
]. On the opposite residence condition, the shape of COVID-19 vulnerability was estimated based on a random infection of 10% of the population living in private households (excluding individuals living in retirement homes) of 81 countries [27
]. In this study, it was estimated that national age and coresidence patterns can alter the vulnerability of a country to COVID-19 outbreaks, with direct effects dependent on a country’s age structure and indirect effects dependent on the size and age structure of a country’s households [27
The specific comorbidities that emerged as risk factors for mortality in our study (Cardiac disease, Kidney disorder, and Neuromuscular disorder) should be interpreted considering the COVID-19 physiopathology. In our study, the prevalence of Kidney disease on admission in patients with COVID-19 was high and associated with clinical stage decline. The highest in-hospital mortality rate (24.4%) and chances of mortality (2.43-fold increase) were registered for individuals with Kidney disease precondition. Other studies in patients with Kidney disorders recorded comparable results. A meta-analysis of eleven COVID-19 studies registered an association between acute Kidney injury and a higher risk of mortality of almost 16-fold (OR = 15.93), with creatinine levels significantly higher in non-survivors compared to survivors [28
]. Nevertheless, the meta-analysis reported a high heterogeneity and a difficulty in adjusting for confounders [28
]. Pathophysiological mechanisms may be involved. Considering that CO2 is an independent determinant of pH adjusted by alveolar ventilation, the disturbance of the acid–base regulation (through the interplay of bicarbonate buffer and respiratory and renal systems) may induce acid–base imbalance and in this way may pose a life-threatening situation [29
]. A second hypothesis consists in the direct Kidney infection by SARS-CoV-2, which recognizes the human Angiotensin-converting enzyme 2 as a cellular receptor that allows it to infect different host cells, a mechanism previously expressed by SARS-CoV virus [30
]. This mechanism could explain the particular importance of acute Kidney injury during hospitalization, considering the exhibited conditions of proteinuria; hematuria; and elevated levels of either serum creatinine, blood urea nitrogen, or both, rendering a significant increase of in-hospital mortality between 1+ and 5.5-fold [31
Patients with cardiovascular disorder exhibited a nearly 3-fold increase in the chance of dying from COVID-19. Cardiovascular disease has been consistently reported as one of the main risk factors for COVID-19 mortality. Two recent meta-analysis reported an odds ratio (95% CI) of 5.19 (3.25; 8.29) (1) and a risk ratio (95% CI) of 2.25 (1.53; 3.29)) for mortality in cardiovascular disease patients [32
]. Our results are in concordance with the systematic review by Pranata et al. [32
] and lower than the systematic review of Zheng et al. [15
]. A possible reason may be the small study effect on the estimates from Zheng et al. [15
], considering the majority of the 13 studies included were of smaller sample size compared to both our study and Pranata et al. [32
]. Nevertheless, the direction of the estimate towards risk is clear. The pathophysiological mechanisms behind this association may be multiple: from severe infection with SARS-CoV-2, precipitating myocardial infarction, myocarditis, heart failure, and arrhythmias as well as an acute respiratory distress syndrome and renal failure [33
]; through the evolution along with multiorgan failure directly due to SARS-CoV-2-infected endothelial cells and resulting endothelitis [33
]; to the potential impacts of therapies considering the likely increase in the number of ACE 2 receptors and the corresponding increase in the susceptibility [33
]. Furthermore, the link between pneumonia and cardiovascular complications should be accounted for: recent studies explore/registered myocardial injury during SARS-CoV-2, secondary to type 2 myocardial infarction, a consequence of increased oxygen demand or reduced oxygen supply during respiratory failure [34
]. In this scenario, cytokines microvascular activation can cause not only myocardial injury but also harm other systems involved in COVID-19 infections, including the kidneys [37
]. Consequently, the relation between cardiovascular disease, Kidney disease and Diabetes should not be ruled out [33
Patients with Neuromuscular disorders registered a 41% increase in lethality. The pathophysiological mechanisms related to this association could be: (i) the fact that patients with this precondition are under the use of immunosuppressive therapies and therefore more likely to increase the severity of COVID-19 infection [39
]; and (ii) risk of exacerbation of myasthenia gravis and QT prolongation in patients with pre-existing cardiac involvement secondary to the treatment with hydroxychloroquine and azithromycin [40
]. However, given that the present database did not provide the patient-specific pharmacological therapies, this question remains open.
Diabetes, previously registered as a significant risk factor for COVID-19 mortality [3
], was not significant in our study when adjusted for other variables of interest. A potential reason for this result might be related to the level of glycemic control. A recent study evaluating the impact of blood glucose control and outcomes of COVID-19 in pre-existing type 2 diabetes noted that when adjusting the model for well-controlled blood glucose, a marked lower mortality was registered compared to individuals with poorly controlled blood glucose [41
]. Data from a national study in Portugal with 1688 individuals reported about 82% of Portuguese diabetics were pharmacologically medicated and that 60.7% were controlled [42
]. However, since we were not able to retrieve the level of glycemic control in diabetic individuals in our study, this question remains open.
The strengths of this study include being a population-based study, the large sample size, and the origin of the data. SINAVE is the electronic platform for notification and cases monitoring of mandatory communicable diseases, allowing the analysis and evaluation of emergent situations, particularly large-scale epidemic outbreaks and pandemics, such as a COVID-19. This system allows for the electronic articulation of doctors (who notify cases of illness), health authorities (responsible for epidemiology at local, regional and national levels), and laboratories (cases notification and cases confirmation). SINAVE allows real-time notification, admitting the implementation of control measures to control and limit the spread of disease and the occurrence of additional cases. The SINAVE database is relevant for the present COVID-19 study by its quality and extensibility since it is based on the information registered by the Medical Doctor who notified the case. Furthermore, it contains all the notified cases up to the extraction of the data due to the interoperability characteristics between the SINAVE computer application and clinical process computer applications.
The limitations of the present study include the absence of potentially important data from the database, missing data, underreporting of mild cases, the impossibility of accounting for the temporal sequence of events, and under-reporting of preconditions. The limitations are presented and discussed in detail below, including the potential bias and corresponding direction. The database did not include reported symptoms and laboratory test results. The existence of unknown values in the data for some outcomes, together with the reporting of preconditions in the medical record, may lead to an underestimation of some risk indicators. It is likely that some of the preconditions were under-reported both in quantity and importance: a particular example is cerebrovascular disease, considered one of the comorbidities with significant impact in COVID-19 prognostic [7
], and conspicuously absent from our database. The temporal sequence of events was not taken into account (time elapsed between the onset of symptoms and hospital admission, or time between hospital admission and death), which may imply an underestimation of preconditions. Finally, because the patients’ clinical observation is still ongoing, many individuals have not reached clinical endpoints (recovery or death). The authors performed complementary calculations to evaluate data robustness and model consistency: The results of these complementary estimations (one model excluding cases from the last 3 weeks and the other model on hospitalized cases) show reasonable consistency with the full multivariable model. A further limitation is related to the data that concern only the initial phase of the pandemic in Portugal up to 21 April. The pandemic is still ongoing, registering on June 25 about 40,866 infected individuals and 1555 deaths [43
]. Nevertheless, no update of the data was made available up to the moment, and therefore the hypothesis of a difference in mortality between the initial and actual phase remains open.
The results of the present study registered potential different pathophysiological mechanisms for COVID-19 mortality, suggesting the need for a team approach between different medical specialties in order to maximize the probabilities of recovery for COVID-19 patients. Future research with larger data sets should include the study of effect and impact of preconditions with individuals reaching clinical endpoints to gain a better understanding of risk factors, as well as the economic and health impacts of COVID-19.