Introduction

The COVID-19 pandemic has raised numerous challenges in managing hospitalized patients1. Given the paucity of specific antiviral therapy for COVID-19, supportive care has been a cornerstone of clinical management, with supplemental oxygen and mechanical ventilation being critical interventions for severe cases2,3. Even though COVID-19 is a viral disease, the empirical use of antibacterial agents is very common4,5. Bacterial co-infections are known to complicate viral respiratory illnesses, contribute to increased morbidity and mortality, and require prompt antibacterial therapy6,7. While bacterial co-infection rates in severe influenza can reach 20–30%, the prevalence and characteristics of such infections in COVID-19 patients are less well characterized7,8,9. Differentiating between COVID-19 and bacterial infections can be challenging due to overlapping symptoms and imaging findings, especially with limited resources10,11.

Current guidelines for community-acquired pneumonia recommend initial empirical antibiotic treatment due to the frequent coexistence of bacterial infections despite the lack of definitive diagnostic tests at the onset of pneumonia12,13. Preliminary studies suggest that antibiotics were prescribed in over 70% of COVID-19 cases, primarily based on suspicion of bacterial co-infection. However, emerging evidence suggests that actual rates of bacterial co-infection among hospitalized individuals with COVID-19 may be below 15%, with some studies reporting even lower figures13,14,15. The exact role of antibiotics in this respect is still unclear.

Furthermore, overuse of antibiotics in COVID-19 pneumonia can increase anti-microbial resistance and lead to complications such as Clostridium difficile infections and renal failure. Therefore, antibiotic decisions should be based on the risk of multi-drug-resistant bacteria and potential complications14,16,17.

The importance of rational antibiotic use cannot be overstated, as the irresponsible or incorrect use of these antimicrobials can lead to serious consequences. Overprescribing or inappropriate use of antibiotics can increase healthcare costs and contribute to the rise in the risk of drug toxicities and adverse drug interactions, thereby compromising patient safety18.

Understanding antibiotic prescribing patterns in COVID-19 may help improve the quality and safety of antibiotic use. This study aims to address the current knowledge gap by evaluating antibiotic use, identifying factors influencing antibiotic use, and assessing factors related to mortality in patients hospitalized with mild-to-severe COVID-19.

Materials and methods

Study design and setting

This retrospective cohort study enrolled patients who were hospitalized with a COVID-19 diagnosis at Bitlis Tatvan State and Kastamonu Training and Research Hospital between December 1, 2020, and June 1, 2022. A total of 445 patients met the inclusion criteria, which comprised those with SARS-CoV-2 detected by PCR (n = 431, 97%) or those with clinically compatible signs and symptoms (n = 14, 3%), bilateral pulmonary infiltrates, or high clinical suspicion lymphopenia. The bed occupancy rate was 93%. The Clinical Research Ethics Committee of KTO Karatay University approved this study (2022/019; 23.05.2022). Due to the retrospective nature of the study informed consent was waived by the KTO Karatay Ethics Committee and anonymous clinical data were utilized in the analysis. The research has been performed in accordance with the Declaration of Helsinki.

Participants

The preliminary and confirmed diagnoses of COVID-19 pneumonia, along with all treatment approaches, were determined following the guidelines established by the Ministry of Health Scientific Committee19. During the study period, all patients who were hospitalized with COVID-19 pneumonia were evaluated for eligibility. The inclusion criteria were as follows: the patient was 18 years or older, had positive nasal or nasopharyngeal RT-PCR test results or strong CT findings suggestive of COVID-19 pneumonia, and had not received antibiotics or had received antibiotic therapy within 24 h of admission. Eligible patients were included in the study if they were hospitalized. The study excluded subjects aged < 18 years or > 89 years, pregnant or lactating women, active malignancy, immunosuppressed subjects, and patients receiving additional antibiotic therapy for at least 24 h after hospital admission.

Data collection

A standard data collection form was used to collect demographic, clinical, laboratory, and radiological data from the electronic medical records (Sisohbys, Hospital Information Management System, Turkey). Only one result such as on antibiotic usage and biochemical parameters per patient was included in the study.

The presence of hypertension, history of smoking, chronic obstructive pulmonary disease (COPD), chronic cardiac conditions, diabetes mellitus, oral corticosteroid therapy during hospitalization, tumor presence, and immunosuppression was recorded.

During their hospitalization, from admission to discharge or death, patients' clinical symptoms and laboratory results were typically monitored as follows: Laboratory tests included routine blood tests such as complete blood count and serum biochemistry (including lactate dehydrogenase [LDH], creatinine, C-reactive protein [CRP], procalcitonin, ferritin, and D-dimer). CT scans were performed on all hospitalized patients. The criteria for determining disease severity on admission were based on the COVID-19 diagnosis and treatment protocols issued by the Turkish Ministry of Health19. The severity of infection was classified according to the NIH COVID-19 Treatment Guidelines at the time of hospital admission20. In addition, patient comorbidities and the Charlson Comorbidity Index (CCI), which estimates the risk of mortality, were documented21,22,23.

Antibiotic usage and other drugs

All medications taken during their hospital stay and COVID-19 vaccine information were also recorded. Medications prescribed for COVID-19 included favipiravir, lopinavir/ritonavir, corticosteroids, hydroxychloroquine, monoclonal antibodies (tocilizumab, anakinra), Intravenous immunoglobulin therapy (IVIG), COPD medications, antihypertensives, antidiabetics, antiarrhythmics, and antidepressant/antipsychotic drugs. The use of antibiotics in all patients who met the inclusion criteria was compared between survivors and non-survivors according to the severity of COVID-19. Classes of antibiotics prescribed (β-lactams, second- and third-generation cephalosporins, fluoroquinolones, glycopeptides, linezolid, colistin), time of antibiotic initiation (on admission or empiric vs post-admission), duration of treatment, antibiotic administration, the prevalence of bacterial co-infections (blood, respiratory and urinary tract) and antibiotic use rates were recorded. At the beginning of the first wave, studies reporting the efficacy of azithromycin in combination with hydroxychloroquine as a potential treatment for COVID-19 led to its widespread use, and azithromycin was prescribed both for this purpose and its antibacterial properties.

To determine the rate of bacterial co-infections among these patients, all positive microbiology results and suspected or culture-confirmed bacterial co-infections documented by the physician in clinical records after COVID-19 diagnosis were collected.

Statistical analysis

All statistical analyses were conducted using the IBM SPSS software, version 23 (IBM Corp., Armonk, N.Y., USA). Categorical variables are presented as frequencies (n) and percentages (%), while continuous variables are displayed as medians with interquartile ranges or means ± standard deviations (SD). Categorical variables were analyzed using the chi-square test or Fisher's exact test. The Kolmogorov–Smirnov test was used to confirm the normality assumption for continuous variables. For the comparison of independent continuous variables between survivors and non-survivors, the student's t-test or the Mann–Whitney U-test was utilized, depending on whether or not the statistical hypotheses were met. Repeated measures analysis of variance was conducted to compare differences between and within groups for time-dependent variables. A univariate and multivariate logistic regression model was employed to evaluate the association between antibacterial therapy and clinical mortality. For survival analysis, the multivariable Cox model was used to analyze the variables that were significant in comparing survival and non-survival variables. Odds ratios (ORs), Hazard ratios (HR), and 95% confidence intervals (CIs) were calculated and reported. The level of statistical significance for all tests was set at < 0.05.

Results

Patients and clinical characteristics with antibiotic usage

A total of 445 patients hospitalized with COVID-19 disease were included in the study. The mean age of the patients was 57 ± 18 years (range, 18–89 years). The mean age of patients receiving antibiotics was 60 ± 16 years, which was higher than that of those not receiving antibiotics, at 45 ± 18 years (p < 0.001). Of the patients, 274 (62%) were male and 171 (38%) were female. Nasal or nasopharyngeal RT-PCR tests for SARS-CoV-2 were confirmed positive in 431 (97%) of patients. The remaining 14 patients had bilateral lung infiltrates on their CT scans and a high clinical suspicion for COVID-19. The demographic and clinical profiles of the patients are detailed in Table 1.

Table 1 Patient demographics and clinical characteristics between with antibiotic treatment (n = 354) and without antibiotic treatment (n = 91).
Full size table

Patients were categorized into two groups based on whether or not they received antibiotic therapy. Antibiotics were administered in 354 cases (80%), while antibiotics were not used in 91 cases (20%).

Of all the patients, the most common symptom was sore throat (65%), followed by cough (57%) and fatigue (54%). In addition to these symptoms, myalgia, headache, arthralgia, nausea, and flutter were higher in antibiotic-treated patients (p < 0.05). Comorbidities were present in 279 (63%) patients. Compared to the without antibiotics group, comorbidities and the associated medications were higher in the antibiotic-treated group (p < 0.05, Table 1).

Clinical evaluation, medications, and laboratory results with antibiotic usage

During the follow-up period, 93 patients (21%) were admitted to the intensive care unit. Of these, 47 patients (11%) required invasive mechanical ventilation and 46 patients (10%) required non-invasive mechanical ventilation. Although the necessity for ICU admission varied among groups, the duration of ICU stays showed no differences. The mean hospital stay was 14 ± 10 days. The duration of hospital stays in the group receiving antibiotics was longer than the group not receiving antibiotics (p < 0.001).

Clinical evaluation scores for patients receiving and not receiving antibiotics were evaluated. The CCI was 3 (IQR, 1–5) in the antibiotic group and 1 (IQR, 0–3) in the non-antibiotic group (p < 0.001). Oxygen saturation (SO2) levels of the patients were evaluated during their hospital stay, on the day of admission for antibiotic treatment, and at the time of discharge. The groups receiving antibiotics had lower SO2 values upon admission (p < 0.001).

During the follow-up period, low- and high-dose corticosteroids were more commonly used in those receiving antibiotics, while hydroxychloroquine was more commonly used in those not receiving antibiotics. Furthermore, lopinavir/ritonavir (4%), tocilizumab/anakinra (3%), and IVIG (3%) were used only by the antibiotic group.

CT findings of 428 patients were evaluated during hospitalization. Pulmonary infiltration was detected in 192 (45%) of 428 patients; 151 patients received antibiotic treatment. Regarding in-hospital mortality, 55 out of 56 patients were in the antibiotic group, and one was in the non-antibiotic group (p < 0.001; Table 2).

Table 2 The comparison of clinical evaluation and medications between with antibiotic treatment (n = 354) and without antibiotic treatment (n = 91).
Full size table

Cultures were taken from 298 patients and 61 samples from 38 patients were positive (%13). 59 of the 61 positive cultures were seen in patients receiving antibiotics. The positive cultures were found in 26 blood, 15 sputum, and 20 urine samples. These cultures were accepted as microbiologically confirmed bacterial co-infection.

Table 3 presents the laboratory parameters with antibiotic usage. When laboratory parameters were compared between the antibiotic and without antibiotic groups, neutrophil, CRP (C-reactive protein), and ferritin levels were higher in the antibiotic group (Table 3).

Table 3 The comparison of laboratory parameters between with antibiotic treatment (n = 354) and without antibiotic treatment (n = 91) at admission.
Full size table

Clinical characteristics and antibiotic usage of survivors and non-survivors

Table 4 present the survivors and non-survivors clinical characteristics and antibiotic usage. The survival rate among all patients was 87%. Non-survivors (69 ± 12) were older than survivors (55 ± 18; p < 0.001). In addition, the proportion of non-survivors was substantially higher in males (%80).

Table 4 The comparison of patient demographics and clinical characteristics between the survivors (n = 389) and the non-survivors (n = 56).
Full size table

As defined by NIH guidelines20, 445 COVID-19 patients were categorized into three groups according to the severity of their disease: mild (42% (n = 186), moderate (19% (n = 85), and severe (39% (n = 174). All mild COVID-19 patients were included in the survivors.

The mean length of hospital stay was 14 ± 10 days. The length of hospital stay was longer in non-survivors (21 ± 13) than in survivors (13 ± 8). In terms of ICU admission was higher among non-survivors (62%). The mean length of ICU stay was also higher among non-survivors (18 ± 13) than survivors (8 ± 10; p = 0.001).

During follow-up, 354 patients were treated with antibiotics for 12 ± 7 days. The duration of antibiotic treatment was 18 ± 9 days in the non-survivors, while it was 10 ± 5 days in the survivors (p < 0.001).

While 299 (77%) of the 389 survivors were treated with antibiotics, 55 of the 56 non-survivors were treated with antibiotics. The most frequently used antibiotics were the fluoroquinolones, followed by the beta-lactam antibiotics. In addition, a combination of glycopeptide, linezolid, and colistin was also given. Within the fluoroquinolones, moxifloxacin (n = 117), levofloxacin (n = 88), and ciprofloxacin (n = 6) were used. Of the remaining patients, 17 used glycopeptides or linezolid, and 16 used colistin. The distribution of antibiotics prescribed is given in Table 4.

Univariable and multivariable logistic regression analysis influencing antibiotic use in patients with COVID-19

Table 5 presents the univariable and multivariable logistic regression analysis influencing antibiotic use in patients with COVID-19. Logistic regression analysis was used to determine factors associated with antibiotic use. Variables significant (p < 0.05) or borderline significant in other analyses were included in univariate logistic regression analysis (p < 0.10).

Table 5 Presents the results of the univariate and multivariate logistic regression models, analyzing the factors influencing antibiotic use in patients with COVID-19.
Full size table

Univariate regression analysis was performed on eight variables: age, sex, pneumonic infiltration, SO2 on admission, ICU admission, CCI, corticosteroid use, CRP, and neutrophil count on admission. The multivariate regression analysis included the variables found to be statistically significant in the univariate regression analysis and we found that CCI (OR 1.6 [95% CI 1.1–2.3]), CRP levels (OR 2.3 [95% CI 1–5.1]), and ICU admissions (OR 8.8 [95% CI 1.1–71.3]) influence antibiotic prescriptions (Table 5).

We also examined the Cox regression analysis to assess the factors related to in-hospital mortality in mild to severe COVID-19 disease. However, we did not find an association between antibiotic use and mortality (HR 2.7 [95% CI 0.4–20] (Table 6), but there was just one death in the non-antibiotics group.

Table 6 Multivariate Cox regression model for mortality in patients with COVID-19.
Full size table

Discussion

A key challenge during the COVID-19 pandemic has been the lack of substantial evidence of reliable treatment options24,25. Due to the virus's ability to change rapidly and the newness of the disease, the uncertainty of proven and effective treatments has led to challenges in clinical management26. Although COVID-19 is a viral disease, antibiotics may have to be prescribed in several situations, such as a concurrent clinical suspicion of bacterial pneumonia, patients with multiple comorbidities, or elevated markers of inflammation27. These factors raise concerns about bacterial co-infections and may influence decisions to prescribe antibiotics as a precautionary measure to prevent potential bacterial complications.

This retrospective cohort study demonstrated a high rate of antibiotic prescriptions in hospitalized COVID-19 patients from two health centers in Turkey. Approximately 80% of the patients included in the study received antibiotics during their hospitalization. Despite the high rate of antibiotic prescriptions, microbiologically confirmed bacterial co-infections were relatively low in our study, with only 14% of patients being affected. The bacterial co-infection rates vary from one country to another and even in different communities within the same country, ranging from 3 to 53% in COVID-19 patients25,28,29,30,31,32,33,34.

A study conducted in the United States showed a high antibiotic prescribing rate, reaching 83% of patients who received at least one course of antibiotics despite a low rate of microbiologically confirmed infection (12%)30. In a retrospective study of 1269 COVID-19 patients in 2020–2022 in two Italian hospitals, 84% of patients (n = 1062) received antibiotic treatment, with only 15% having an obvious source of bacterial infection31. A multicentre point-prevalence study conducted in Turkey with a large participant population showed that the antibiotic prescription rate was 75%. However, the rate of clinically or microbiologically confirmed bacterial infections was 29%, and culture positivity was 7%25. In our study, the relationship between antibiotic prescription rate and microbiologically confirmed bacterial co-infection was consistent with previous studies.

Two systematic reviews and meta-analyses by Langford et al. reported the prevalence of antibiotic prescription to be 72% (95% CI 56 to 88%) and 75% (95% CI 68–80%), and the prevalence of bacterial co-infection to be 7% (95% CI 4–10%) and 9% (95% CI 5–15%), respectively32,33. The high antibiotic prescribing rate compared to the low bacterial co-infection rate in our study and other studies may be related to empirical antibiotic use and the severe clinical conditions of the patients.

In early studies, clinicians followed the initiation of treatment for COVID-19 patients based on local guidelines for community-acquired pneumonia12. In addition, some medical centers have recommended empiric antibiotics for the majority of COVID-19 patients based on institutional guidelines30. The most commonly used antibiotics in the treatment of COVID-19 patients are fluoroquinolones, macrolides, and beta-lactams25,30,33. In our study, respiratory fluoroquinolones such as levofloxacin, moxifloxacin, and ciprofloxacin were the most prescribed antibiotics, followed by beta-lactams, especially second/third-generation cephalosporins. These results suggest that healthcare providers prefer these antibiotics to treat respiratory tract infections, especially in COVID-19 patients. In addition, beta-lactams, fluoroquinolones, and their combinations were more commonly used in survivors. In contrast, high-end antibiotics such as glycopeptide/linezolid and colistin were more commonly used in non-survivors (p < 0.001). This highlights the importance of disease severity and specific medical interventions to guide antibiotic prescribing decisions in critically ill patients.

Understanding the predictive factors for the need for antibiotic treatment not only ensures proper management of antibiotic therapy but also improves the patient's overall prognosis and enhances the management of antimicrobial resistance (AMR)33.

The patient's comorbidities, clinical symptoms, and laboratory findings were evaluated to determine whether or not they had received antibiotic treatment. The lack of precise data on the relationship between comorbidity and antibiotic use in COVID-19 may have led to a perception that antibiotics were prescribed at a higher rate in patients without comorbidities. Al-Hadidi et al. reported no difference in antibiotic use between patients with and without comorbidities34. In contrast to this study, we observed that comorbidities and the use of medications associated with comorbidities were higher in the group receiving antibiotics. Calderón-Parra et al. concluded that fewer comorbidities, dry cough, and flu-like symptoms may be associated with inappropriate antibiotic use14. In our study, flu-like symptoms such as sore throat, cough, fatigue, myalgia, arthralgia, headache, and nausea were more common in patients treated with antibiotics.

Laboratory abnormalities have contributed to the use of antibiotics in COVID-19 patients. A recent meta-analysis indicated that procalcitonin (PCT) has limited predictive value for detecting co-infection in patients with COVID-19. However, lower PCT levels appear to be associated with a reduced probability of co-infection35.

Although studies have shown that elevated CRP levels are associated with an increased frequency of antibiotic prescription, CRP is not a reliable indicator of antibiotic prescription during the COVID-19 pandemic. This may be due to the respiratory distress leading to elevated CRP levels observed in the initial presentation of patients with COVID-1925,36,37,38. In our study, we observed that elevated CRP levels in patients who received antibiotics contributed to increased antibiotic use. However, PCT (procalcitonin) values could not be assessed for all patients due to limited laboratory resources.

In our study, we used univariate logistic regression analysis to determine factors associated with antibiotic use. We identified eight independent factors that were associated with an increased use of antibiotics. These factors are as follows: age, gender, Charson comorbidity index, requirement for supplemental oxygen, steroid usage, presence of moderate/diffuse lung involvement, neutrophil count, and C-reactive protein levels. These factors significantly influenced the higher antibiotic utilization among the individuals included in the study (Table 5).

A study in Turkey identified several risk factors associated with antibiotic use in COVID-19 patients, including age, hospitalization in ICU, need for supplemental oxygen, moderate or diffuse lung involvement, and a lymphocyte count < 800. It was also reported that moderate or extensive lung involvement and CRP levels above the upper limit of normal (ULT) were associated with antibiotic use25. In a meta-analysis study conducted on 30,212 patients, it was observed that antibiotic prescribing prevalence was more common in older age groups and in patients with higher severity of illness33.

In our study, the in-hospital mortality rate was 12.6%, which is in line with the results of other studies in the literature31. Mortality was observed in 14.1% of the patients treated with antibiotics. The group of patients receiving antibiotics consisted of clinically more severe cases with high pulmonary infiltration, low SO2 saturation and high comorbidities. Thus, the severity of illness was also a cause for administering antibiotics, creating a selection bias between the two groups. We also wanted to examine the effect of antibiotics on mortality; however, due to the presence of selection bias, this analysis could not be effectively evaluated. Although our study did not find an association between antibiotic use and increased mortality, we observed that patients treated with antibiotics had higher rates of comorbidities and severe forms of COVID-19.

Some limitations should be recognized for this study. The retrospective nature of this study, the small number of patients, and the fact that it was a two-center study conducted predominantly in the Turkish population may limit the generalizability of the results. In addition to the fact that similar microbiologic tests were performed in patients and only the results of positive tests were reported, there is also a lack of data on isolated bacteria and their antibiotic susceptibility. Furthermore, bacterial co-infections may have been missed due to antibiotic treatment before hospitalization, and microbiology sampling may have influenced microbiology test results. The presence of an unknown false-negative rate for cultivation. Another limitation is that it was not possible to measure PCT levels in all patients due to laboratory problems, and the duration of MV support is unknown. Although our study has several limitations, its strength is that it covers both the quarantine period and the period when no strict pandemic measures were taken (December 2020 to June 2022).

Conclusion

In summary, our study reveals a concerning trend of considerable antibiotic use in COVID-19 patients, even in the absence of clinical or laboratory diagnosis. This practice may have serious consequences due to the low rate of bacterial co-infection and the potential risks associated with antibiotic use, such as the development of antimicrobial resistance and drug-induced toxicity. Therefore, it is crucial to limit the empirical use of antibiotics in COVID-19 patients and to consider antibiotic treatment only when there is clear microbiologic evidence or strong clinical suspicion of bacterial infection. These precautions minimize the potential harm of unnecessary antimicrobial use while preserving the efficacy of these important drugs for future medical needs. Further research is necessary to more accurately identify patients with bacterial co-infections who would benefit from antibiotic treatment.