Appropriateness of empiric antibiotic therapy for community-acquired pneumonia: A multicenter cross-sectional study

Abstract

Background

Community-acquired pneumonia (CAP) remains a major public health concern, with appropriate empiric antibiotic therapy critical to improving patient outcomes and combating antimicrobial resistance (AMR). Despite the availability of national guidelines, adherence among physicians in Lebanon was previously unclear.

Objective

To evaluate the appropriateness of empiric antibiotic prescribing for hospitalized CAP patients in Lebanon and identify factors associated with non-compliance with national guidelines.

Methods

A cross-sectional study was conducted in five Lebanese hospitals across various regions. Medical records of 380 adult CAP patients were reviewed. Empiric antibiotic regimens were evaluated for compliance with national CAP guidelines based on drug selection, dosage, route, and duration. Multivariate logistic regression was used to identify predictors of inappropriate prescribing.

Results

Only 38.9 % of patients received fully guideline-concordant empiric therapy. Inappropriate prescribing was significantly associated with physician rank (residents: aOR = 29.13, p = 0.001), geographic region (South Lebanon: aOR = 7.78, p = 0.028), and antibiotic class (β-lactam plus respiratory fluoroquinolones: aOR = 19.18, p < 0.001). Clinical factors such as systemic inflammatory response syndrome and elevated serum creatinine were also linked to inappropriate use.

Conclusion

Non-compliance with national CAP treatment guidelines is widespread in Lebanese hospitals and driven by physician experience, geographic disparities, and misalignment in antibiotic selection. Strengthening antimicrobial stewardship programs, regional training, and clinical decision support is essential to improve adherence and reduce AMR risk.

1. Introduction

Community-acquired pneumonia (CAP) is one of the most prevalent and serious infectious diseases worldwide, especially among the elderly (≥65 years) and patients with underlying comorbidities.^1,2 It remains a leading cause of hospitalization and mortality, with the World Health Organization estimating approximately 4 million deaths annually due to CAP.^1,3 In the Middle East and North Africa region, lower respiratory tract infections account for approximately 10 % of mortality, underscoring the clinical and public health burden posed by CAP.⁴

Despite advances in diagnosis and supportive care, clinical outcomes for hospitalized CAP patients remain suboptimal. Thirty-day mortality ranges from 5 % to 15 %, rising to 20–30 % in cases requiring intensive care.^5,6 Numerous factors influence these outcomes, including patient characteristics, disease severity, and the timeliness and appropriateness of antibiotic therapy.7, 8, 9, 10 Guideline-concordant empiric antibiotic treatment has been associated with significantly reduced mortality and improved clinical outcomes.¹¹ Conversely, inappropriate prescribing—whether due to delayed initiation, selection of ineffective agents, or use of unnecessarily broad-spectrum antibiotics—has been linked to treatment failure, prolonged hospitalization, and increased rates of readmission and mechanical ventilation.^10,12,13

Inappropriate antibiotic use also plays a central role in the global antimicrobial resistance (AMR) crisis. Overuse and misuse of antibiotics, especially in hospital settings, have led to the emergence of multidrug-resistant organisms and an increase in healthcare-associated complications such as Clostridioides difficile infections.14, 15, 16, 17 In patients with CAP, infections caused by resistant pathogens are associated with twice the mortality and financial burden compared to those caused by susceptible strains.^18,19 The economic impact is substantial, with hospitalization costs reaching over $20,000 per patient due to complications and extended care requirements.^19,20

Empirical antibiotic therapy remains the cornerstone of CAP treatment, as microbiological confirmation is often delayed. However, early broad-spectrum coverage must be balanced with the need to minimize resistance by adjusting therapy once pathogen susceptibility is known.^1,21 National and international guidelines emphasize this balance and promote antibiotic stewardship as a critical strategy to ensure optimal outcomes and limit the spread of resistance.

In Lebanon, the Lebanese Society of Infectious Diseases and Clinical Microbiology (LSIDCM) has issued national guidelines for CAP management, adapted from global standards and informed by local resistance data.²² These guidelines recommend empiric antibiotic selection based on disease severity, comorbidities, and regional pathogen susceptibility patterns.^21,23,24 Nonetheless, antibiotic prescribing practices remain suboptimal in Lebanon, with rising resistance rates and limited data on adherence to these guidelines in real-world hospital settings.²⁵

Previous research in Lebanon has already highlighted challenges with guideline adherence. For instance, a multicenter retrospective study conducted in the emergency departments of three Lebanese hospitals found that adherence to guidelines for all community-acquired infections was only 32.6 %.²⁵ While this study provided valuable initial insights, its scope included all community-acquired infections within the emergency department setting. To date, data specifically evaluating empiric antibiotic guideline adherence for hospitalized CAP patients, a population with higher disease severity and complexity, remain limited on a national level. Our study builds upon this foundational work by providing the first nationwide analysis focused exclusively on empiric antibiotic therapy for hospitalized CAP patients, thereby offering a more detailed assessment of prescribing practices within the inpatient setting and clarifying the factors that drive non-compliance in this high-risk group.

Given the significant consequences of inappropriate prescribing and the growing threat of AMR, evaluating adherence with national CAP treatment guidelines is essential. This study aims to assess the appropriateness of antibiotic prescribing for hospitalized CAP patients in Lebanon and to identify factors associated with deviations from guideline-recommended care.

2. Methods

2.1. Study design and setting

This cross-sectional study involved the analysis of patients' medical records. The participants were drawn from five university hospitals located in various regions of Lebanon: two in Beirut and one each in the South, North, and Mount Lebanon. These sites represent main tertiary care and teaching centers within their respective geographic regions. This sampling strategy ensures the study sample reflects broader clinical practices across the country, thereby enhancing the generalizability of the findings beyond a single institution or region.

2.2. Study population

This study included adult patients (18 years and older) of both sexes who were hospitalized for CAP with accessible medical records. We excluded patients with healthcare-associated, hospital-associated, or ventilator-associated pneumonia (since they are not included in CAP guidelines), as well as outpatients, individuals with concurrent infections, and populations not covered by national guideline recommendations (such as children and pregnant women).

2.3. Sources of data

Data collection was retrospective review of medical records for patients admitted between January 2024 and June 2024. Patient medical records, identified using ICD-10 code J18.9, were evaluated against eligibility criteria. The collected data included demographics, clinical presentation (signs and symptoms), comorbidities, laboratory results, culture findings, and antibiograms. We also documented antibiotic prescriptions, hospital stay length, and clinical outcomes (resolution of symptoms, complications, or mortality).

Antibiotic regimens were assessed for appropriateness according to Lebanese national CAP management guidelines,²² considering antibiotic choice, dosage, route of administration, and treatment duration. A regimen was classified as overall appropriate only when all four aspects met guideline recommendations. To provide a clear framework for this evaluation, a summary of the guideline's main recommendations is available in the Supplementary Material.

Patient outcomes, including resolution of symptoms or clinical improvement, deterioration leading to ICU (intensive care unit) admission, and death were analyzed to evaluate treatment efficacy and safety.

2.4. Microbiological analysis and susceptibility testing

Microbiological data were collected from sputum and blood cultures when performed as part of standard clinical care. Laboratory analysis focused on identifying common bacterial pathogens associated with CAP, including Streptococcus pneumoniae, Haemophilus influenzae, Pseudomonas aeruginosa, and Klebsiella pneumoniae. Antimicrobial susceptibility testing was performed and interpreted according to the Clinical and Laboratory Standards Institute (CLSI) guidelines. The panel of tested antibiotics included key classes recommended in national CAP guidelines, such as third-generation cephalosporins, piperacillin/tazobactam, carbapenems, macrolides, and respiratory fluoroquinolones (RFQ).

2.5. Study size

The sample size was determined using Epi Info software to achieve a 95 % confidence interval with a 5 % allowable margin of error. Consequently, the minimum required sample size calculated was 384 participants. To obtain the study sample, a simple random sampling method was used. From the list of all patients admitted with CAP at each of the five participating hospitals during the study period, 100 medical records were randomly selected for screening. In total, 500 medical records were initially analyzed in this review (100 patient records from each hospital), with 120 being excluded (60 were children, 10 were pregnant women, 20 had multiple infections, and 30 had hospital and ventilator-associated pneumonia). After assessing the eligibility of the participants, a total of 380 individuals were included in the current study.

2.6. Statistical methods

Data analysis was performed using IBM SPSS® software version 27. Descriptive statistics were presented as frequencies and percentages for categorical variables and as means with standard deviations (SD) for continuous variables. Normal distribution of continuous variables was assessed through visual inspection of histograms and verification that skewness and kurtosis absolute values were less than 1.96.

For bivariate analyses, associations between categorical variables were evaluated using the Pearson's chi-square test (χ²) or Fisher's Exact test when appropriate. Continuous variables were compared using Student's t-test for normally distributed data or Mann-Whitney U test for non-parametric data.

Variable selection for multivariate analysis was based on the disjunctive cause criterion.²⁶

This criterion helps select potential confounding variables by including any factor that is a cause of either the outcome (inappropriate prescribing) or a key exposure. A multivariate logistic regression model was employed to identify factors associated with inappropriate antibiotic prescribing, using the appropriateness of empirically prescribed antibiotic therapy as the dependent variable. The model incorporated patient demographics, comorbidities, physician characteristics, CAP symptoms, laboratory parameters, and empiric antibiotic regimens as covariates. The model was optimized using backward stepwise elimination, a method that refines the model by starting with all potential variables and systematically removing the least influential ones until only statistically significant predictors (p < 0.05) remain.

Adjusted odds ratios (aOR) with corresponding 95 % confidence intervals (CIs) were calculated to quantify the strength and direction of associations between independent variables and outcomes. Statistical significance was established at p < 0.05, with 95 % CIs that did not include 1.00 considered statistically significant.

2.7. Ethical consideration

The study obtained ethical approval through a comprehensive process to protect patient confidentiality and welfare. The Institutional Review Board (IRB) at Beirut Arab University granted approval under code 2023-H-0092-P-R-0547, ensuring compliance with established ethical standards. Additionally, the ethical committees of the participating hospitals reviewed and approved the study, demonstrating a commitment to ethical guidelines and respectful access to medical records. To further ensure confidentiality, patient identifiers were removed during data collection and analysis.

3. Results

3.1. Patient demographics, comorbidities, and physician characteristics

A total of 380 hospitalized patients with CAP were included in this study. The mean age was 67.98 ± 17.01 years, with an approximately equal gender distribution. The majority of patients (67.9 %) were admitted to internal medicine departments, with 42.4 % of admissions occurring in Beirut-based facilities. Most patients (92.1 %) had medical coverage. Prevalent comorbidities included hypertension (55.0 %), diabetes mellitus (36.3 %), and dyslipidemia (26.3 %). Approximately one-third of patients were current smokers. The mean length of hospital stay was 7.52 ± 4.02 days. Prescribing physicians were predominantly male (84.7 %) and specialized in pulmonary medicine (47.4 %), with most being consultants or specialists (51.1 %). Detailed demographic data are presented in Table 1.

Table 1.

Demographics, Comorbidities, Clinical Presentation, Lab Data, and Physician Characteristics in CAP Hospitalized Patients (Total = 380).

Parameter	Frequency (%) or mean ± SD	Parameter	Frequency (%) or mean ± SD
Age	67.98 ± 17.01	Length of hospital stay	7.52 ± 4.02
Gender		Physician Gender
Male	188 (49.5 %)	Male	322 (84.7 %)
Female	192 (50.5 %)	Female	58 (15.3 %)
Admission Department		Physician Position
Internal medicine	258 (67.9 %)	Consultant or Specialist	194 (51.1 %)
Critical care	122 (32.1 %)	Attending or Fellow	129 (33.9 %)
Hospital Area		Resident	57 (15 %)
Beirut	161 (42.4 %)	Physician Specialty
Mount Lebanon	75 (19.7 %)	Pulmonary	180 (47.4 %)
North Lebanon	105 (27.6 %)	Infectious Diseases	82 (21.6 %)
South Lebanon	39 (10.3 %)	Internal Medicine	87 (22.9 %)
Marital Status		Others (Emergency, Cardiology)	31 (8.2 %)
Single	54 (14.2 %)	Clinical Presentation
Married	295 (77.6 %)	Fever /chills	213 (56.1 %)
Divorced/ widow/widower	31 (8.2 %)	Cough	242 (63.7 %)
Medical Coverage		Dyspnea	287 (75.5 %)
Yes	350 (92.1 %)	Pleuritic chest pain	74 (19.5 %)
No	30 (7.9 %)	Confusion	130 (34.2 %)
Smoking Status		Wheezing	186 (48.9 %)
Current smokers	123 (32.4 %)	Decreased breath sounds	163 (42.9 %)
Non-smokers	257 (67.6 %)	Rales /crackles	161 (42.4 %)
Comorbid Medical Conditions		Tachypnea	128 (33.7 %)
Diabetes Mellitus	138 (36.3 %)	SIRS	189 (49.7 %)
Hypertension	209 (55 %)	Vital Signs and Other Laboratory Findings
Dyslipidemia	100 (26.3 %)	Temperature (° C)	38.41 ± 1.04
Congestive heart failure	44 (11.6 %)	Heart rate (beats/ min)	87.19 ± 16.05
Coronary artery disease	69 (18.2 %)	Respiratory rate (breaths/min)	22.51 ± 5.75
Cerebrovascular disease	15 (3.9 %)	SBP (mm Hg)	125.17 ± 21.63
Neoplastic disease	18 (4.7 %)	DBP (mm Hg)	71.61 ± 12.59
Renal Impairment	57 (15 %)	WBC *103 (cells/ μL)	12.21 ± 4.56
CNS Disorders	47 (12.4 %)	Neutrophils (%)	75.15 ± 15.71
Respiratory disease	77 (20.3 %)	Serum Creatinine (mg/dL)	1.13 ± 0.87
Allergy to antibiotics	5 (1.3 %)	BUN (mg/dL)	30.14 ± 18.09
Antibiotics use within 3 months*	27 (7.1 %)	Oxygen saturation (%)	92.24 ± 7.01
Living in long-term care facility	29 (7.6 %)	Hematocrit (%)	33.08 ± 7.05
		2	262 (68.9 %)
CURB-65 Score		3	97 (25.5 %)
		4	20 (5.3 %)
		5	1 (0.3 %)

BUN: blood urea nitrogen - CNS: central nervous system – DBP: diastolic blood pressure- SBP: systolic blood pressure- SIRS: systemic inflammatory response syndrome- WBC: white blood cells *Antibiotic use within 3 months: The use of any systemic antibiotic for any clinical indication within the 90 days prior to the current hospital admission.

3.2. Clinical presentation and laboratory findings

The clinical manifestation of CAP was characterized primarily by respiratory symptoms (Table 1), with dyspnea (75.5 %), cough (63.7 %), and fever/chills (56.1 %) being most frequently observed. Physical examination findings and laboratory parameters indicated moderate disease severity, with 68.9 % of patients having CURB-65 scores of 2. CURB-65 is the clinical tool used to assess CAP severity and help decide on hospitalization. It stands for:

• oC: Confusion
• oU: Urea level > 20 mg/dL
• oR: Respiratory rate ≥ 30 breaths per minute
• oB: Blood pressure (systolic <90 mmHg or diastolic ≤60 mmHg)
• o65: Age ≥ 65 years

3.3. Microbiological findings and antibiotic susceptibility

Microbiological sampling was performed in a subset of patients, including sputum cultures (21.8 %) and blood cultures (32.1 %). As detailed in Table 2, the predominant pathogen isolated was Streptococcus pneumoniae (44.3 %). Other microbes were identified whereas the remaining cultures yielded no growth.

Table 2.

Microbiological Findings and Overall Antibiotic Susceptibility Rates.

A. Distribution of Isolated Pathogens
Pathogen	Frequency (%)
Streptococcus pneumoniae	44.3
Candida species	12.3
Haemophilus influenzae	9.8
Pseudomonas aeruginosa	6.5
Klebsiella pneumoniae	3.3
Acinetobacter baumannii	2.5
B. Overall Antibiotic Susceptibility Rates of Bacterial Isolates
Antibiotic Class	Susceptibility (%)
Carbapenems	61.4
Third-generation cephalosporins	49.4
Macrolides	37.1
Respiratory fluoroquinolones	35.5
β-lactams
Piperacillin/tazobactam	54.2
Aztreonam	44.6

Antimicrobial susceptibility testing revealed the highest rates of sensitivity to carbapenems (61.4 %), while comparatively lower susceptibility was observed for macrolides (37.1 %) and RFQ (35.5 %). Based on these culture results, empiric antibiotic therapy was subsequently de-escalated in 55.7 % of cases.

3.4. Empiric antibiotic prescribing patterns and guideline-based appropriateness assessment

The empiric antibiotic regimens listed in Table 3 reflect the prescribing patterns observed in this study and are not presented in a specific order of recommendation. According to the Lebanese national guidelines, the primary determinant for empiric antibiotic selection is patient severity, which dictates the site of care. For non-critically ill patients hospitalized on a regular ward, the first-line recommendation is a β-lactam plus a macrolide. For critically ill patients admitted to the ICU, treatment is further stratified based on risk factors for Pseudomonas aeruginosa infection, which warrants broader-spectrum antimicrobial coverage. Alternative regimens, such as RFQs, are explicitly reserved as a second choice for non-ICU patients, primarily for cases of penicillin allergy or intolerance. The guidelines actively discourage their first-line use in this population to mitigate the risk of promoting AMR.

Table 3.

Empiric Antibiotics Prescribed for CAP Patients, Prescription Appropriateness, and Patient Outcomes in the Study Sample.

Empiric Antibiotics	Frequency total = 380	Percentage %
β-lactam + macrolide	92	24.2
β-lactam +RFQ	158	41.6
RFQ alone (Moxifloxacin, Levofloxacin)	43	11.3
Antipseudomonal β-lactam + RFQ	65	17.1
Antipseudomonal β-lactam + RFQ + vancomycin	18	4.7
Antipseudomonal β-lactam + macrolide +AG	4	1.1
Carbapenems included	50	13.2
Appropriateness of empiric antibiotic prescription
Appropriate selection	153	40.3
Appropriate dosage	313	82.4
Appropriate route of administration	372	97.89
Appropriate duration	350	92.11
Overall appropriateness of antibiotic prescriptions	148	38.9
Patient outcomes
Resolution of symptoms/clinical improvement	336	88.4
ICU admission	22	5.8
Death	22	5.8

AG: aminoglycoside, ICU: intensive care unit, RFQ: respiratory fluoroquinolones

Among the CAP patients, β-lactam plus RFQ was the predominant empiric antibiotic regimen (41.6 %), followed by β-lactam plus macrolide (24.2 %). The overall appropriateness of antibiotic prescriptions was only 38.9 %, despite high rates of appropriate route of administration (97.89 %) and duration (92.11 %).

The primary concern identified was inappropriate antibiotic selection, with only 40.3 % of prescriptions deemed appropriate. Despite the fact that the standard empiric antibiotic regimen for patients admitted to regular wards consists of a combination of β-lactam and macrolide agents, the most frequently prescribed therapy was β-lactam plus RFQ combination, which is primarily recommended for ICU patients according to national guidelines. Additionally, RFQ monotherapy was inappropriately utilized as first-line treatment in non-ICU settings, whereas guidelines recommend reserving it as an alternative for patients with penicillin allergy.

Inappropriate dosing was another significant issue, largely attributable to failure to adjust dosages of renally eliminated antibiotics—particularly levofloxacin, imipenem/cilastatin, and meropenem—in patients with renal impairment. Notwithstanding the observed prescribing discordance with national guidelines, favorable clinical outcomes were documented in 88.4 % of patients. Detailed antibiotic prescribing patterns, appropriateness metrics, and patient outcomes are demonstrated in Table 3.

3.5. Bivariate analysis: factors contributing to inappropriate empiric antibiotic prescribing

The bivariate analysis of factors linked to inappropriate antibiotic prescribing in hospitalized CAP patients revealed several significant associations as presented in Table 4. Significant geographic variations in inappropriate antibiotic prescribing were observed across Lebanon, with South Lebanon having the highest rate (79.5 %) compared to Beirut (47.2 %, p < 0.001). Internal medicine departments showed more inappropriate prescribing than critical care units (66.3 % vs 50 %, p = 0.002). Patients with dyslipidemia and hypertension experienced lower inappropriate prescribing (44 % and 56.9 % respectively, p < 0.001 and p = 0.035), while renal impairment was associated with higher inappropriate prescribing (77.2 %, p = 0.007). Clinical presentations associated with appropriate prescribing included confusion, cough, dyspnea, fever, and physical findings like decreased breath sounds and tachypnea (all p < 0.01). β-lactam plus macrolide combinations demonstrated the highest appropriate prescribing rate (92.4 %), while RFQ alone showed the lowest (7 %, p < 0.001). Patients with appropriate prescriptions exhibited higher vital signs (temperature, respiratory rate, pulse) but lower BUN and creatinine levels (all p < 0.05). The presence of SIRS was also associated with more appropriate prescribing (46.6 % vs 31.4 %, p = 0.002).

Table 4.

Bivariate analysis of factors linked to inappropriate antibiotic prescribing in hospitalized patients with CAP:(total = 380).

Variable	Appropriate Prescriptions N (%) or mean ± SD Total = 148	Inappropriate Prescriptions N (%) or mean ± SD Total = 232	Total	p-value
Hospital Area
Beirut	85 (52.8 %)	76 (47.2 %)	161	<0.001⁎
North Lebanon	34 (32.4 %)	71 (67.6 %)	105
South Lebanon	8 (20.5 %)	31 (79.5 %)	39
Mount Lebanon	21 (28 %)	54 (72 %)	75
Hospital Department
Internal medicine	87 (33.7 %)	171 (66.3 %)	258	0.002⁎
Critical care units	61 (50 %)	61 (50 %)	122
Physician gender
Male	122 (37.9 %)	200 (62.1 %)	322	0.318
Female	26 (44.8 %)	32 (55.2 %)	58
Physician position
Consultant / Specialist	104 (53.6 %)	90 (46.4 %)	194	0.185
Attending	32 (24.8 %)	97 (75.2 %)	129
Resident	12 (21.1 %)	45 (78.9 %)	57
Physician specialty
Pulmonary	67 (37.2 %)	113 (62.8 %)	180	0.157
Internal medicine	27 (31 %)	60 (69 %)	87
Infectious diseases	46 (56.1 %)	36 (43.9 %)	82
Others (Emergency medicine, Cardiology)	8 (25.8 %)	23 (74.2 %)	31
Patients Data
Age	67.70 ± 15.44	68.16 ± 17.97	–	0.212
Length of hospital stay	7.80 ± 3.63	7.34 ± 4.25	–	0.276
Gender
Male	64 (34 %)	124 (66 %)	188	0.052
Female	84 (43.8 %)	108 (56.2 %)	192
Medical coverage
No	13 (43.3 %)	17 (56.7 %)	30	0.608
Yes	135 (38.6 %)	215 (61.4 %)	350
Smoking
No	96 (37.4 %)	161 (62.6 %)	257	0.357
Yes	52 (42.3 %)	71 (57.7 %)	123
Diabetes mellitus
No	97 (40.1 %)	145 (59.9 %)	242	0.304
Yes	51 (37 %)	87 (63 %)	138
Hypertension
No	58 (33.9 %)	113 (66.1 %)	171	0.035⁎
Yes	90 (43.1 %)	119 (56.9 %)	209
Dyslipidemia
No	92 (32.9 %)	188 (67.1 %)	280	<0.001⁎
Yes	56 (56 %)	44 (44 %)	100
Coronary artery disease
No	120 (38.6 %)	191 (61.4 %)	311	0.963
Yes	28 (40.6 %)	41 (59.4 %)	69
Congestive heart failure
No	131 (39 %)	205 (61 %)	336	0.807
Yes	17 (38.6 %)	27 (61.4 %)	44
Cerebrovascular disease
No	144 (39.5 %)	221 (60.5 %)	365	0.421
Yes	4 (26.7 %)	11 (73.3 %)	15
CNS disorders
No	127 (38.1 %)	206 (61.9 %)	333	0.515
Yes	21 (44.7 %)	26 (55.3 %)	47
Neoplastic disease
No	140 (38.7 %)	222 (61.3 %)	362	0.665
Yes	8 (44.4 %)	10 (55.6 %)	18
Respiratory disease
No	116 (38.3 %)	187 (61.7 %)	303	0.805
Yes	32 (41.6 %)	45 (58.4 %)	77
Renal impairment
No	135 (41.8 %)	188 (58.2 %)	323	0.007⁎
Yes	13 (22.8 %)	44 (77.2 %)	57
Allergy
No	145 (38.7 %)	230 (61.3 %)	375	0.383
Yes	3 (60 %)	2 (40 %)	5
Antibiotic use within the past 3 months
No	135 (38.2 %)	218 (61.8 %)	353	0.309
Yes	13 (48.1 %)	14 (51.9 %)	27
Living in long-term care facility
No	137 (39 %)	214 (61 %)	351	0.898
Yes	11 (37.9 %)	18 (62.1 %)	29
Clinical Symptoms and Physical Examination
Confusion
No	71 (28.4 %)	179 (71.6 %)	250	<0.001⁎
Yes	77 (59.2 %)	53 (40.8 %)	130
Cough
No	41 (29.7 %)	97 (70.3 %)	138	0.005⁎
Yes	107 (44.2 %)	135 (55.8 %)	242
Dyspnea
No	23 (24.7 %)	70 (75.3 %)	93	0.001⁎
Yes	125 (43.6 %)	162 (56.4 %)	287
Fever
No	52 (31.1 %)	115 (68.9 %)	167	0.006⁎
Yes	96 (45.1 %)	117 (54.9 %)	213
Pleuritic Chest Pain
No	108 (35.3 %)	198 (64.7|%)	306	0.003⁎
Yes	40 (54.1 %)	34 (45.9 %)	74
Decreased Breath Sounds
No	65 (30 %)	152 (70 %)	217	0.006⁎
Yes	83 (50.9 %)	80 (49.1 %)	163
Wheezing
No	53 (27.3 %)	141 (72.7 %)	194	0.001⁎
Yes	95 (51.1 %)	91 (48.9 %)	186
Rales/Crackles
No	70 (32 %)	149 (68 %)	219	0.002⁎
Yes	78 (48.4 %)	83 (51.6 %)	161
Tachypnea
No	66 (26.2 %)	186 (73.8 %)	252	<0.001⁎
Yes	82 (64 %)	46 (36 %)	128
CURB-65 Score
2	82 (31.3 %)	180 (68.7 %)	262	<0.001⁎
3	57 (58.8 %)	40 (41.2 %)	97
4	8 (40 %)	12 (60 %)	20
5	1 (100 %)	0 (0 %)	1
Empiric Antibiotic Class
β-lactam + macrolide	85 (92.4 %)	7 (7.6 %)	92	<0.001⁎
β-lactam +RFQ	22 (13.9 %)	136 (86.1 %)	158
RFQ alone	3 (7 %)	40 (93 %)	43
Antipseudomonal β-lactam + RFQ	36 (55.4 %)	29 (44.6 %)	65
Antipseudomonal β-lactam + RFQ + vancomycin	1 (5.6 %)	17 (94.4 %)	18
Antipseudomonal β-lactam + macrolide +AG	1 (25 %)	3 (75 %)	4
Carbapenems Included
No	137 (41.5 %)	193 (58.5 %)	330	0.008⁎
Yes	11 (22 %)	39 (78 %)	50
Vital Signs and Laboratory Findings
Temperature (°C)	38.19 ± 1.03	37.89 ± 1.03	–	0.005⁎
Respiratory rate (breaths/min)	23.61 ± 6.17	21.63 ± 5.25	–	0.044⁎
Pulse (beats/min)	90.02 ± 14.79	84.86 ± 16.70	–	0.001⁎
SBP (mmHg)	126.10 ± 21.68	124.41 ± 21.64	–	0.596
DBP (mm Hg)	72.02 ± 13.08	71.28 ± 12.22	–	0.913
WBC ⁎103 (cells/ μL)	12.17 ± 4.68	11.91 ± 4.64	–	0.610
Neutrophils (%)	75.10 ± 15.68	75.18 ± 15.78	–	0.966
BUN (mg/dL)	26.38 ± 14.88	32.58 ± 19.56	–	0.005⁎
Serum Creatinine (mg/dL)	0.97 ± 0.51	1.24 ± 1.02	–	0.001⁎
Hematocrit (%)	37.11 ± 7.14	35.54 ± 6.96	–	0.052
Oxygen saturation (%)	91.52 ± 7.02	92.83 ± 6.98	–	0.139
SIRS
No	60 (31.4 %)	131 (68.6 %)	191	0.002⁎
Yes	88 (46.6 %)	101 (53.4 %)	189

BUN: Blood Urea Nitrogen, CNS: central nervous system, DBP: diastolic blood pressure, RFQ: respiratory fluoroquinolones, SBP: systolic blood pressure, SIRS: systemic inflammatory response syndrome, WBC: white blood cells

^⁎Denotes a significant p-value <0.05.

3.6. Multivariable logistic regression analysis of factors influencing inappropriate empiric antibiotic prescribing

The multivariable logistic regression analysis identified key factors associated with inappropriate empiric antibiotic prescribing for hospitalized CAP patients. Junior physicians demonstrated significantly higher rates of inappropriate prescribing, with residents showing the highest risk (aOR = 29.13, p = 0.001), followed by attending physicians (aOR = 12.54, p = 0.002), compared to consultants/specialists. Regional differences were evident, with South Lebanon showing a significant increase in inappropriate prescribing compared to Beirut (aOR = 7.78, p = 0.028). Regarding antibiotic choices, all alternatives to the recommended β-lactam+Macrolide combination (in non critically ill patients) showed significantly higher odds of inappropriate use, particularly β-lactam+RFQ (aOR = 19.18, p < 0.001) and RFQ alone (aOR = 15.29, p < 0.001). Clinical presentations with wheezing and higher pulse were associated with more appropriate prescribing. Notably, patients with a CURB-65 score 3 had 98 % lower odds of receiving inappropriate antibiotics (aOR = 0.02, p < 0.001). Elevated creatinine levels also tripled the risk of inappropriate prescribing (aOR = 3.83, p = 0.045). Detailed results of the multivariable analysis are presented in Table 5.

Table 5.

Factors Influencing Inappropriate Empiric Antibiotic Prescribing in Hospitalized CAP Patients: Multivariable Logistic Regression Analysis.

Factor	Wald	Adjusted Odds Ratio(aOR)	95 % Confidence Interval		p-value
Physician position (Reference: Consultant/Specialist)
Attending	10.07	12.54	2.63	59.76	0.002⁎
Resident	10.30	29.13	1.55	28.25	0.001⁎
Hospital Area (Reference: Beirut)
South Lebanon	4.85	7.78	1.25	48.24	0.028⁎
Empiric antibiotic prescribed (Reference: B-Lactam + Macrolide)
β-lactam + RFQ	8.52	19.18	5.36	63.25	<0.001⁎
RFQ alone	6.84	15.29	3.64	47.05	<0.001⁎
Antipseudomonal β-lactam + RFQ	4.25	6.59	1.35	25.14	<0.001⁎
Antipseudomonal β-lactam + RFQ + vancomycin	2.64	9.09	2.59	38.19	<0.001⁎
Carbapenem included	5.98	8.13	1.52	15.59	0.015⁎
Clinical and Lab Data:
Wheezing	9.68	0.15	0.05	0.49	0.002⁎
Pulse (beats/min)	5.16	0.96	0.92	0.99	0.023⁎
Respiratory rate (breaths/min)	5.02	0.89	0.81	0.98	0.025⁎
SIRS	10.19	18.43	3.08	70.27	0.001⁎
Serum Creatinine (mg/dL)	3.83	1.94	0.99	3.78	0.048⁎
CURB-65 score 3	25.92	0.02	0.00	0.08	<0.001⁎

RFQ: respiratory fluoroquinolones, SIRS: systemic inflammatory response syndrome.

-Omnibus test of model coefficients p-value <0.001.

-Hosmer and Lemeshow test p-value = 0.807.

-Model summary: Nagelkerke R² = 0.803.

-Classification table overall percentage = 90.5 %.

-constant p-value = 0.701.

^⁎Denotes a significant p-value <0.05.

4. Discussion

This multicenter, cross-sectional study provides the first nationwide analysis of empiric antibiotic prescribing for hospitalized CAP patients in Lebanon, revealing a significant gap between clinical practice and national guidelines. A key finding is the low overall adherence rate of just 38.9 %, indicating that the majority of patients did not receive fully guideline-concordant empiric therapy. This rate is concerning and aligns with previous research in Lebanon that found a 32.6 % adherence rate for all community-acquired infections in emergency departments.²⁵ When benchmarked internationally, our findings are comparable to rates in Latin America (30.6 %) and Denmark (31 %) but fall well below those reported in Italy (56.7 %) and Norway (80 %), highlighting a substantial opportunity for improvement.27, 28, 29, 30

Inappropriate antibiotic selection was the primary driver of non-adherence to treatment guidelines. The most frequently prescribed regimen was a β-lactam combined with RFQ, accounting for 41.6 % of cases. In contrast, the guideline-recommended combination of a β-lactam plus a macrolide was utilized in only 24.2 % of prescriptions. This prescribing pattern directly contradicts national guidelines,²² which reserve RFQs as an alternative agent for specific indications, such as severe infections in the ICU. This is due to the strong potential of RFQ to induce AMR in Lebanon,^31,32 and these recommendations aimed to decrease collateral damage and contain antibiotic resistance trends on the national level.²²

The multivariable analysis quantified this deviation, revealing that prescribing a β-lactam plus an RFQ or an RFQ alone was associated with a 19-fold and 15-fold increased odds of inappropriate use, respectively. This widespread overuse of broader-spectrum agents underscores a critical disconnect between guideline recommendations and clinical practice, which likely contributes to rising AMR trends. These findings are consistent with previous reports; for instance, a Lebanese medical review identified antibiotic selection as the most common reason for guideline non-adherence, aligning with a Latin American study where unnecessary broad-spectrum antibiotic use (40.0 %) was a major factor for non-adherence.^25,28 These data highlight an urgent need for targeted educational interventions and antimicrobial stewardship programs (ASPs) to improve guideline adherence and mitigate AMR crisis.

Our analysis also identified significant variations in prescribing practices linked to both physician seniority and hospital location.

The non-adherence of physicians to clinical guidelines is a multifaceted concern that undermines patient care and increases healthcare costs, a finding consistent with a substantial body of literature. Although recommendations for empiric therapy in CAP are straightforward, clinicians often deviate from standard protocols, with international studies reporting suboptimal rates of appropriate prescribing ranging from 32.6 % to 61.5 %.^18,25,33 The drivers of this gap are complex, encompassing both physician-related and systemic factors. A particularly significant physician-related factor is the level of clinical experience; a clear gradient was observed whereby residents and attending physicians demonstrated a markedly higher likelihood of inappropriate prescribing compared to consultants and specialists. This may be ascribed to variances in clinical dynamics, as less experienced physicians manage a broader range of conditions and may struggle to consistently internalize specific guidelines, a vulnerability previously identified in the literature.³⁴ This aligns with other individual factors, including knowledge gaps and the tendency to prescribe more aggressively based on a subjective assessment of a patient's severity. Compounding these issues are systemic barriers, such as practical obstacles to implementing timely de-escalation in busy clinical settings and external pressures from pharmaceutical marketing.³⁵ This context underscores that the mere existence of guidelines is insufficient, highlighting the critical need for robust ASPs, especially those providing targeted support for less experienced clinicians, to address these intertwined barriers.

Furthermore, significant geographic disparities emerged, with hospitals in South Lebanon showing a nearly eight-fold increase in inappropriate prescribing compared to those in Beirut. This urban-rural gap likely reflects differences in access to a specialized workforce, resources, advanced infrastructure, and proximity to research and education, which are typically concentrated in capital cities.³⁶

The relationship between a patient's clinical presentation and prescribing appropriateness was notably complex. The multivariable analysis indicated that specific, localized signs of a lower respiratory tract infection, such as wheezing accompanied by elevated pulse and respiratory rates, were associated with a decrease in inappropriate prescribing. This occurs because these objective signs create a clear clinical picture of bacterial pneumonia, increasing diagnostic certainty and prompting physicians to adhere closely to the Lebanese guidelines for CAP, which outline a specific, evidence-based therapeutic pathway. In contrast, the presence of a generalized SIRS was associated with higher rates of inappropriate prescribing. While SIRS signals severe illness, it is a non-specific syndrome that lacks diagnostic precision and can be triggered by non-infectious causes. This combination of high perceived severity and low diagnostic certainty often compels clinicians to adopt an aggressive, defensive approach with broad-spectrum antibiotics. This practice deviates from the core principle of targeted therapy emphasized in Lebanese antimicrobial stewardship guidelines, thus leading to an increase in prescribing that is classified as inappropriate.

Interestingly, certain markers of higher disease severity were associated with more appropriate prescribing. For instance, patients with a CURB-65 score of 3 had 98 % lower odds of receiving inappropriate therapy. Patients with such scores or clinical presentations exhibiting severe disease are typically admitted to the ICU, requiring intensive monitoring and support.²² This finding can be explained by the structure of the national guidelines, which provide highly specific, protocol-driven regimens for severe CAP cases requiring ICU admission. Specialists managing these critically ill patients are likely more familiar with and adherent to these clear-cut protocols.³⁷

Conversely, the presence of renal impairment emerged as a robust predictor for inappropriate prescribing, tripling the risk. This association is primarily attributable to the failure to modify dosages for antibiotics that undergo renal elimination, notably specific β-lactams and fluoroquinolones.³⁸ This is consistent with established pharmacokinetic principles, as compromised renal function significantly alters drug clearance, creating a risk of systemic accumulation and potential toxicity if dosing regimens are not adjusted accordingly.³⁸ The magnitude of this issue is well-documented in the literature, with reports indicating that between 9.7 % and 53.1 % of antibiotic prescriptions for patients with renal insufficiency are dosed inappropriately.^39,40 Furthermore, antibiotics represent the most frequently implicated drug class, comprising 39.8 % of all medications that necessitate dose adjustment.³⁹ A significant gap between experience and confidence may contribute to this problem, as one study found that while 95 % of physicians have experience prescribing for patients in renal failure, 35 % reported lacking confidence in accurate dose determination.⁴¹

Our study also highlighted the significant underutilization of antibiotic de-escalation, a core principle of ASP. Empiric therapy was narrowed in only 55.7 % of cases for which culture results were available. This rate is consistent with international findings, such as a study in eight Venezuelan hospitals reporting a 57 % de-escalation rate, and reflects a low transition to targeted therapy also observed in Norway, indicating that barriers to this practice are widespread.^30,42

De-escalation, the timely modification of initial empiric therapy based on culture and susceptibility results, is an essential component of modern ASPs. Its implementation offers well-established advantages, including reducing the selective pressure that drives AMR, minimizing collateral damage such as Clostridioides difficile infections, and decreasing hospital stays and healthcare costs.⁴³ Crucially, the safety of this practice is well-supported. A large observational study of 11,159 patients with community-acquired pneumonia found that de-escalation was non-inferior to continued broad-spectrum therapy regarding both 15-day and in-hospital mortality.⁴⁴ A primary contributing factor to the low adherence observed in our cohort is that the current Lebanese national guidelines lack a formal, structured protocol for de-escalation. This represents a significant gap and a missed opportunity to optimize antibiotic use. Incorporating a clear, evidence-based de-escalation protocol into future national guidelines is therefore a critical step toward improving clinical practice and combating the growing threat of AMR.

Our findings provide a clear roadmap for developing tailored ASPs, though successful implementation hinges on a realistic and detailed assessment of the feasibility, potential barriers, and strategic future directions within the Lebanese healthcare context. The significant non-adherence among junior physicians highlights the urgent need to integrate stewardship principles directly into residency training curricula. This intervention is highly feasible by leveraging existing partnerships between the LSIDCM, universities, and teaching hospitals. However, the primary barriers include the already crowded nature of medical curricula, which would require displacing other content, and the challenge of gaining buy-in from program directors who must be convinced of its priority. A strategic future direction to overcome these hurdles is the development of a standardized, case-based e-learning module on CAP management, which offers a flexible and scalable solution that can be integrated without major curricular overhauls. Beyond education, systemic safeguards are needed to address issues like the overuse of RFQs and failures in renal dose adjustment. Implementing Clinical Decision Support (CDS) systems within electronic health records (EHRs) presents a moderately feasible technological solution, but only in the growing number of hospitals that have adopted such systems. The barriers to this approach are substantial, encompassing not only the high initial cost and ongoing maintenance but also the technical challenge posed by the heterogeneity of EHR systems across institutions, which complicates a unified rollout. Furthermore, the risk of alert fatigue, whereby clinicians become desensitized and ignore frequent notifications, could render a poorly designed system ineffective. To navigate these complexities, a future pathway involves piloting a CDS tool in a large university hospital to gather crucial data on designing effective, non-intrusive alerts and to demonstrate a return on investment that could encourage wider adoption. Finally, to address the striking geographic disparities and promote consistent practices like de-escalation, a national audit and feedback program is a highly feasible intervention. The primary obstacles are not technical but organizational and political, rooted in the need for robust data sharing agreements and the potential for institutional resistance if the program is perceived as punitive rather than supportive. Building trust through guaranteed data confidentiality is paramount. Therefore, the most effective future direction is to launch a voluntary pilot program, initially targeting underserved regions, to demonstrate its value in guiding local quality improvement and to build momentum for a comprehensive national system focused on tracking key performance indicators like de-escalation rates and inappropriate RFQ use.

This study has several limitations. The retrospective observational design relied on medical records, which may vary in completeness and detail, potentially introducing information bias and limiting the ability to infer causality. However, essential variables were sufficiently documented to support robust analysis.

Although the study included hospitals from diverse regions, the Beqaa region was not represented. Nonetheless, previous studies^33,45 have reported similarly high rates of inappropriate prescribing across Lebanon, suggesting the findings remain broadly generalizable.

The binary classification of guideline adherence may oversimplify complex clinical decisions, as some deviations may have been clinically justified but still categorized as non-adherent. Additionally, residual confounding cannot be fully excluded, as physician-specific factors (e.g., prescribing habits, years of experience, specialty training) and institutional variables were not captured.

Despite these limitations, this is the first nationwide study in Lebanon to evaluate empiric antibiotic prescribing practices for hospitalized CAP patients. It offers critical insights into guideline adherence and lays the groundwork for future stewardship initiatives and research.

5. Conclusions

This study reveals suboptimal adherence with Lebanese national guidelines for CAP treatment, with inappropriate prescribing driven by factors including physician experience, clinical context, and geographic disparities. These practices contribute to the growing threat of AMR, increased healthcare costs, and the risk of adverse patient outcomes. Strengthening antibiotic stewardship is therefore imperative through targeted, actionable interventions. Our findings support the need to integrate stewardship principles into residency curricula, implement clinical decision support tools to guide appropriate antibiotic selection and renal dosing, and establish a national audit and feedback program to standardize care and promote evidence-based practices like de-escalation. Such multidisciplinary efforts are essential to optimize prescribing and mitigate the serious public health challenge of AMR in Lebanon.

CRediT authorship contribution statement

Deema Rahme: Writing – original draft, Software, Methodology, Formal analysis, Data curation, Conceptualization. Hania Nakkash Chmaisse: Writing – review & editing, Validation, Supervision, Conceptualization. Pascale Salameh: Writing – review & editing, Visualization, Validation, Supervision, Conceptualization.

Funding

The study did not receive funds or grants from any source.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material 1

Supplementary material 2