The first case of COVID-19 (Coronavirus Disease 2019) was reported by the Chinese authorities in December 20191 and has spread at breakneck speed, with 103,597,957 cases reported in 192 countries at the time of writing.2 SARS-CoV-2, which causes the disease, is the seventh known coronavirus that infects human cells. The most common symptoms, according to a systematic review and meta-analysis,3 are fever (88.7%), cough (57.6%) and dyspnoea (45.6%), although studies show a range of incidence, and some report fatigue is also a common symptom4,5,6. However, as the infection has spread and research has grown, numerous clinical presentations have been associated, ranging from mild conditions (conjunctivitis,7 anosmia,8 skin manifestations9–12) to severe presentations (septic shock, pulmonary thromboembolism, acute respiratory distress syndrome, acute renal failure and myocardial involvement).13,14 Studies suggest that clinical deterioration, when it occurs, happens around the eighth day of disease,4 and is associated with a poor evolution and requires a thorough initial follow-up in all patients.

Clinical and epidemiological health strategies carried out since the onset of the pandemic differ substantially between countries. In Spain, primary healthcare had to adapt care to the circumstances imposed by the pandemic,15 starting from an initial scenario marked by the lack of diagnostic methods available to primary healthcare and by the problems of physicians' access to patients by prioritizing telephone care to the detriment of face-to-face visits. Studies16–22 have shown that analysis of the sociodemographic and clinical characteristics of patients with a presentation compatible with COVID-19 provides useful information for decision-making, making it essential to develop models predictive of an unfavorable evolution23–26 based on the differences between patients requiring hospitalization and those treated in primary healthcare.

This study was carried out in the first wave of the pandemic in order to combine information to identify primary healthcare patients with a clinical suspicion of COVID-19 who require referral for hospital assessment.

The objective of this study was to identify clinical and demographic factors predictive of hospitalization, comparing a cohort of patients treated by our primary healthcare centre diagnosed with suspected or confirmed COVID-19 who required hospital admission and those who did not.

We included 518 patients: a diagnóstico was confirmed by PCR or serology in 215 (41.5%) while 303 (58.5%) were classified as clinical suspicion of COVID-19 without diagnostic confirmation. Most patients with clinical suspicion were followed on an outpatient basis whereas the percentage of hospitalized patients was high in those with a confirmed diagnóstico (Fig. 1).

Fig. 1.

Hospitalization in clinical suspicion and confirmed diagnosis patients.

(0.06MB).

The characteristics of the population are displayed in Table 1. Patients who required hospital admission were most often male, aged 65–80 years, and obese or overweight, with an AMG of 4, who did not work face-to-face during the study period. No differences were found in terms of country of origin, smoking, alcohol use or drug use (Table 1). Twelve (2.3%) patients died, including three outpatients and 8 hospitalized patients.

Hospitalization was significantly more frequent in patients with fever, chills, general discomfort, haemoptysis, dizziness, dyspnoea, chest pain and vomiting or nausea (Table 2). However, fewer patients with odynophagia and nasal congestion were hospitalized compared with those without. Hospitalization was non-significantly higher in patients with coughing and diarrhea than in those without.

With respect to comorbidities, hospitalization was significantly higher in patients with COPD, diabetes, hypertension, ischemic heart disease, cerebrovascular disease, chronic kidney disease and obesity, among others (Table 3).

The bivariate analysis between hospitalization and medication showed a higher probability of hospitalization in patients receiving any type of treatment compared with those who were not (Table 4).

27% of patients consulted primary healthcare services on the day of symptom onset, 43.3% consulted between the second and fifth days and 29.7% later.

In the logistic regression model, the independent predictive variables for hospitalization were male sex, age, a history of ischemic heart disease and the presence of dyspnoea, haemoptysis and nausea or vomiting. The coefficients obtained in the predictive model and the OR and 95% CI are shown in Table 5.

The reliability of the predictive model was 86.4%. The negative predictive value was 88.7%, and the positive predictive value was 71%. The sensitivity of the model was 48% and the specificity was 95.4%.

In our cohort of primary healthcare patients with a presentation compatible with COVID-19, the risk factors for hospitalization were male sex, older age, a history of ischemic heart disease and dyspnoea, haemoptysis, nausea or vomiting. During the study period, no COVID-19 diagnostic tools were available in primary healthcare. Due to this, at first the diagnóstico was only confirmed in patients hospitalized at some point during the infection and in essential healthcare workers. Subsequently, thanks to increased accessibility to diagnostic techniques and the development of serologies for the detection of antibodies to SARS-CoV-2, the diagnóstico could be confirmed in more patients. This study aimed to evaluate the ability of primary healthcare to diagnose and manage these patients.

Regarding comorbidities, several studies23,30–35 have found cardiovascular risk factors such as diabetes, hypertension, obesity or cardiovascular disease as the most common comorbidities found in patients diagnosed with COVID-19, as well as renal disease or chronic pulmonary disease. Similarly, our study found that diabetes, high blood pressure, COPD, ischemic heart disease and cerebrovascular disease were added risk factors for COVID-19 hospitalization, as were obesity and chronic kidney disease.

With respect to chronic medication, the bivariate analysis showed that patients receiving ACE inhibitors, ARA II or oral antidiabetics were more frequently hospitalized than those who were not, although the differences disappeared in the multivariate analysis. This suggests that these drugs are not a risk factor in themselves but are associated with diseases such as high blood pressure, ischemic heart disease and diabetes for which they are used. Our study, like others,36–38 helps refute the statements made at the beginning of the pandemic that patients receiving ACE inhibitors had a risk of an adverse evolution due to the mechanism of action of the drug and the pathogenesis of the virus.39,40

With respect to chronic corticosteroid treatment, paradoxically, patients who were receiving them, and who had severe comorbidities that could result in an unfavorable evolution, were not significantly more often hospitalized than those who were not. This may be because, contrary to what was initially thought, patients with active SARS-CoV-2 infection may benefit from corticosteroid therapy by potentially mitigating the inflammatory cascade developed by patients with severe disease, as suggested in some studies.41–44

With regard to the clinical manifestations of Covid-19, different studies3–6,30,45 have suggested dyspnoea, cough or fever as the most prevalent symptoms, as well as myalgia or fatigue, among others. In our study, fever, chills, general discomfort, haemoptysis, and dyspnoea resulted in an increased risk of hospitalization compared with patients without these symptoms. However, we also identified nausea and vomiting as risk factors for hospitalization, unlike other studies which did not.3,4,46 On the other hand, and contrary to what is suggested in reports,3–6,30,45 cough was not a risk factor for hospitalization, while nasal congestion and odynophagia seemed to have a slight protective effect against hospitalization.

The mortality rate was 2.3%, including 7.7% of hospitalized patients and 0.7% of outpatients. The outpatients who died were at high clinical risk due to age and comorbidities. This highlights the value of primary healthcare in the adequacy of therapeutic measures and in ensuring the comfort and home care of patients who require it.

The logistic regression model had a sensitivity of 48%, and a specificity of 95.4%. This suggests that this tool helps identify patients who should be referred for hospital care. However, some patients with a negative test also might benefit from hospital care. In this group, the decision on referral should be based on the clinical criteria of the attending family physician. The formula would gain validity if used in another cohort of primary healthcare patients to predict the probability of hospitalization.

This study analyzed a sample of patients attended in primary care, presenting a wide range of clinical manifestations of COVID-19, and not only severe cases, unlike most published studies, which describe hospitalized patients. Also, because it was conducted in the first wave of the pandemic it provides a description of the care provided by Spanish family doctors with very limited access to diagnostic tests for COVID-19. Furthermore, the model developed makes it possible to identify some risk factors that are worth taking into account to guide clinical decisions in the care of patients with COVID-19. We conducted a review in which we talked with every patient included and did not rely only on the medical record.

However, the study has some limitations. As it is a retrospective cohort based on electronic medical records, there is a recording bias, which we tried to minimize by conducting a telephone interview with all patients during the month after patient recruitment was concluded. Despite this limited period there may also have been minimal memory bias. In addition, the sample is of limited size as it is restricted to patients in a single health centre as we prioritized immediacy in conducting the study given the pandemic situation. In addition, it should be taken into account that due to the exceptional situation, many patients with admission criteria remained at home instead. Also, the evolution of the severity in some cases could be influenced by external factors such as the lack of access to hospital resources due to the collapse of the health system, as well as the difficulty in maintaining the isolation or overcrowding of patients in residences.

In the first pandemic wave, 20% of patients treated for COVID-19 in primary healthcare required hospital admission, most during the first seven days after symptom onset. As determined by the logistic regression model, male sex, older age, a history of ischemic heart disease, and dyspnoea, haemoptysis, nausea or vomiting may be factors for a higher risk of an unfavorable evolution during COVID-19, and these patients should be evaluated carefully. Patients with at least one of the above factors, which correlate with a higher hospital admission rate, should receive a closer follow-up in order to early detect when they may benefit from a hospital evaluation based on their clinical evolution.

The study was conducted in accordance with Spanish legislation (Biomedical Research Act, 14/2007, 3 July) and international norms (Helsinki Declaration and Nuremberg Code), guaranteeing the confidentiality of patient data in accordance with Organic Law on the Protection of Personal Data (3/2018, 5 December) and Law 41/2002, of 14 November, regulating patient autonomy and rights and obligations in matters of information and clinical documentation. The study was approved by the research bioethics committee of the Hospital Clínic de Barcelona (Reg. HCB/2020/0616).

This study received a research grant from the Commission de Recerca ad hoc COVID-19 of the Hospital Clínic de Barcelona and the Fundació Clínic per a la Recerca Biomèdica.