Physicians' prediction for the assessment of atypical pathogens in respiratory tract infections

Abstract Background Patients with acute respiratory tract infections are frequently prescribed antimicrobials despite high rates of virus detection. Physicians may overprescribe antimicrobials owing to the concern of bacterial infections, including those because of atypical pathogens. We investigated the accuracy of clinical predictions concerning atypical pathogen infections. Methods We prospectively enrolled adult patients who presented with a fever and cough in outpatient clinics between December 2016 and August 2018. After taking a history and performing physical examinations, physicians predicted the possibility of respiratory infections because of atypical pathogens. Disease probabilities were categorized into 3 grades (high: ≥50%, intermediate: 20% ≥ and <50%, and low: <20%) and were judged by physicians who were taking care of the patients. Confirmation of atypical pathogens was performed by comprehensive molecular analyses of respiratory samples. Results Atypical pathogens were detected in 21 of 210 patients. A close contact history (odds ratio [OR]: 11.4, 95% confidence interval [CI]: 2.4‐53.5) and the presence of pneumonia (OR: 12.9, CI: 4.3‐39.2) were associated with the detections. Atypical pathogens were detected in 32.3% of high‐probability cases (10/31), while atypical pathogens were only detected in 8.8% of intermediate‐probability cases (8/91) and 3.4% of low‐probability cases (3/88) (P < .001). Conclusions The current study indicates that physicians’ predictions were associated with the detection of atypical pathogens; however, overestimation was observed.


| INTRODUC TI ON
Overprescription of antimicrobials has contributed to the increased prevalence of antimicrobial-resistant bacteria. 1,2 To counteract this trend, national action plans have been developed in many countries. 3 The majority of antimicrobial agents, reportedly up to 90%, are prescribed in an outpatient setting. 4 Acute respiratory tract infections are the primary reason for these prescriptions, 5 and antimicrobial agents are prescribed in up to 70% of patients with upper respiratory tract infections or acute bronchitis in outpatients care settings in both Japan 6 and the United States, 7 despite high virus detection rates in these cases. 8,9 Mycoplasma pneumoniae, Chlamydophila pneumonia, and Bordetella pertussis have been recognized as the main causes of bacterial bronchitis. 10 Differentiation between bacteria and viruses is considered to be challenging, and a previous study reported that there were no specific signs or symptoms associated with M pneumoniae infections. 11 Concerns regarding possible bacterial infections may lead physicians to prescribe antimicrobials, so detecting clues to accurately predict these infections may promote antimicrobial stewardship. Recently, diagnostic scoring criteria for considering atypical pathogen infections among adult pneumonia patients were published by the Japanese Respiratory Society (JRS) and are now widely used in Japan. 12 These criteria were developed to support the diagnosis of atypical pathogens among pneumonia patients, but the criteria were not applied to other respiratory infections.
In this study, we investigated the epidemiology and characteristics of atypical pathogen infections in the outpatient clinic setting using comprehensive molecular analyses in order to evaluate physicians' diagnostic predictions and the performance of the JRS criteria for diagnosing atypical pathogen infections.

| MATERIAL S AND ME THODS
This study was a prospective observational design and performed at the ambulatory clinics of two acute care hospitals between December 2016 and August 2018 in Japan. Written informed consent was obtained from all participants in this study. Ethical approval was granted by the Review Board Committee of each hospital.
All patients who met the study inclusion criteria had respiratory samples obtained for a comprehensive molecular examination ( Figure 1). Physicians initially predicted the potential for atypical pathogen infections after obtaining the patient history and conducting a physical examination. Each physician documented their judgment concerning the potential for atypical pathogen infections as 1 of 3 grades (high: ≥50%, intermediate: 20% ≥ and <50%, and low: <20%). All of the judgments were performed subjectively by the physicians taking care of the patients. If trainees examined study patients, all predictions were performed under the instruction of the attending physicians.
If further tests, including antigen testing (influenza antigen testing, pneumococcal urinary antigen testing, legionella urinary antigen testing, Mycoplasma pneumoniae antigen testing, and rapid antigen detection test for group A streptococcus), blood tests, or radiological imaging, were performed, the diagnostic predictions were reevaluated prior to the patient leaving the clinic on the same day. A laboratory test for M pneumoniae using the particle agglutination (PA) antibody was not available at the study hospitals in a single day. The reevaluation of judgments for the potential for atypical pathogen infections was documented by the same grading system as mentioned above. Each reevaluation was performed by each physician who was taking care of the patients without instruction from other physicians, except for trainees; however, strict rules prohibiting talking about patients between the first and second evaluations were not implemented.
Comprehensive molecular examinations were performed at a later date, and the results were made not available to physicians during the evaluation.

| Inclusion and exclusion criteria for patients
All patients who presented at the clinic with both a fever (1 degree higher than their baseline body temperature or a body temperature >37°C) and cough for at least 3 days were enrolled in this study.
Pediatric patients (age <18 years), patients with unstable physical conditions (eg, shock), a history of multiple exacerbations of chronic pulmonary disease, an apparent history or presence of dysphagia, presence of obstructive pneumonia, lung abscess, empyema, healthcare-associated pneumonia or hospital-onset pneumonia referred from other facilities, tuberculosis, nontuberculous mycobacterium lung infections, pneumomycosis, sinusitis, or tonsillitis were excluded from this study. In addition, patients with a history of a fever or cough for more than 21 days or patients without documentation of their physicians' prediction regarding the probability of atypical pathogen infections were also excluded.

| Data collection and physicians' prediction regarding the probability of detecting atypical pathogens
As background data, we collected information on the age, gender, visiting month, comorbidities, close contact with patients confirmed to have atypical pathogen infections, history of preceding antimicrobial use, history of signs and symptoms (rhinorrhea, sputum, severe cough, sore throat, myalgia, arthralgia, diarrhea, and rash),

| Statistical analyses
Categorical and continuous variables were compared using Fisher's exact test and the Mann-Whitney U-test, where appropriate. A multivariable logistic regression model was constructed to identify variables significantly associated with atypical pathogen positivity.
Variables with P-value <.05 in the univariate analyses were included in the multivariable model. A P-value less than .05 was considered significant. All statistical analyses were performed with the SPSS version 20 software program (IBM).

F I G U R E 1
Flowchart of the study process. When additional tests were not ordered by physicians during care for patients, the prediction of the initial evaluation was used as the final prediction for atypical pathogen infections

| Study case selection and detected pathogens
The flowchart describing the case selection process is shown in

| Physicians' predictions regarding the probability of atypical pathogen respiratory tract infections
As shown in Table 2  Note: Categorical data are presented as numbers (proportion, %).
Continuous data are presented as medians with the interquartile range.

| Performance of atypical pathogen diagnostic scoring criteria based on the Japanese guideline
Abbreviations: NPV, negative predictive value; PPV, positive predictive value. The scoring criteria without laboratory testing consisted of factors (a) to (e), and a score ≥3 was considered indicative of atypical pathogen infection. c The scoring criteria with laboratory testing consisted of factors (a) to (f), and a score ≥ 4 was considered indicative of atypical pathogen infection.

TA B L E 3
Performance of the atypical pathogen diagnostic scoring criteria based on the Japanese guidelines a

| D ISCUSS I ON
Using comprehensive molecular analyses for respiratory pathogens, atypical pathogens were confirmed in approximately 10% of all patients with a fever and cough lasting more than 3 days. A history of close contact with other persons with atypical pathogen infection and the presence of pneumonia were factors significantly associated with the detection of atypical pathogens among these patients.
While physicians' predictions were associated with molecular detection rates of atypical pathogens, the overall detection rates were only about half of those predicted by physicians. Diagnostic scoring for atypical pathogen based on the Japanese guideline has high sensitivity and moderate specificity for detecting atypical pathogens in pneumonia patients.
Bacterial infections are generally reported to account for 5%-10% of acute bronchitis cases. 16   Several limitations associated with the present study warrant mention. First, the current study was only conducted in two Japanese teaching hospitals across a period of approximately one and a half years. The epidemiology of atypical pathogens differs based on country, 26 season, [27][28][29] and year. 27,30 Therefore, the findings of the current study may not be generalizable to other settings.
Second, we analyzed M pneumoniae, B pertussis, and C pneumoniae as atypical pathogens. While the detection of M pneumoniae was performed with two molecular assays, B pertussis and C pneumoniae were evaluated only through multiplex polymerase chain reaction assays. In addition, there were 43 patients (22.8%) with preceding antimicrobial use among atypical cases of pathogen-negative respiratory tract infection. Therefore, the presence of false-negative cases in this study cannot be ruled out. 31 Finally, we did not analyze Legionella pneumophila or other atypical pathogens, including B parapertussis, 32 B holmesii, 33 and C psittaci, 34 although they have been rarely identified in Japan. [35][36][37] A comprehensive molecular analysis indicated that atypical pathogens were detected in only 10% of patients presenting with a fever and cough for more than 3 days. While physicians' predictions were associated with the detection of atypical pathogens, overestimation was observed.