Abstract
Introduction Variable treatment exists for children with bacterial pneumonia complications such as pleural effusion and empyema. Subspecialists at an urban academic tertiary children’s hospital created a literature-based diagnosis and management algorithm for complicated pneumonia in children. We proposed that algorithm implementation would reduce use of computed tomography (CT) for diagnosis of pleural infection, thereby decreasing radiation exposure, without increased adverse outcomes.
Materials and Methods A cross-sectional study was undertaken in children (3 months to 20 years old) with principal or secondary diagnosis codes for empyema and/or pleural effusion in conjunction with bacterial pneumonia. Study cohorts consisted of subjects admitted 15 months before (cohort 1, n = 83) and after (cohort 2, n = 87) algorithm implementation. Data were collected using clinical and financial data systems. Imaging studies and procedures were identified using Current Procedural Terminology codes. Statistical analysis included χ2 test, linear and ordinal regression, and analysis of variance.
Results Age (P = 0.56), sex (P = 0.30), diagnoses (P = 0.12), and severity level (P = 0.84) were similar between cohorts. There was a significant decrease in CT use in cohort 2 (cohort 1, 60% vs cohort 2, 17.2%; P = 0.001) and reduction in readmission rate (7.7% vs 0%; P = 0.01) and video-assisted thoracoscopic surgery procedures (44.6% vs 28.7; P = 0.03), without concomitant increases in vancomycin use (34.9% vs 34.5%; P = 0.95) or hospital length of stay (6.4 vs 7.6 days; P = 0.4). Among patients who received video-assisted thoracoscopic surgery drainage (n = 57), there were no significant differences between cohorts in median time from admission to video-assisted thoracoscopic surgery (2 days; P = 0.29) or median duration of chest tube drainage (3 vs 4 days; P = 0.10). There was a statistically nonsignificant trend for higher rate of pathogen identification in cohort 2 (cohort 1, 33% vs cohort 2, 54.1%; P = 0.12); Streptococcus pneumonia was the most commonly identified pathogen in both cohorts (37.5% vs 27%; P = 0.23).
Discussion Implementation of an institutional complicated pneumonia management algorithm reduced CT scan use/radiation exposure, VATS procedures, and readmission rate in children with a diagnosis of pleural infection, without associated increases in length of stay or vancomycin use. This algorithm provides the framework for future prospective quality improvement studies in pediatric patients with complicated pneumonia.
Pneumonia and its complications are among the most frequent causes of hospital admission for the pediatric population; yet there is great variability in strategies for the treatment of otherwise healthy children who develop complications of bacterial pneumonia, including pleural effusion, empyema, necrotizing pneumonia, and lung abscess. At an urban academic tertiary children’s hospital, a group of providers sought to optimize the diagnosis and care of patients with complicated pneumonia. To develop a consensus for management, a multidisciplinary committee (Complicated Pneumonia Committee) was formed, consisting of physician representatives from the Hospitalist, Infectious Diseases, Pulmonary Medicine, Radiology, Surgery, and Emergency Medicine Divisions. This Complicated Pneumonia Committee developed a diagnosis and treatment management algorithm incorporating the best available evidence to achieve the following primary goals within existing hospital structures and systems: (1) minimize morbidity and mortality, (2) decrease radiation exposure, (3) limit unnecessary antibiotic use, and (4) decrease hospital length of stay (LOS). Secondary goals included improved communication between subspecialists and hospitalists, provision of adequate postdischarge follow-up care with the appropriate care providers, and establishment of the framework for future prospective quality improvement studies in pediatric patients with complicated pneumonia.
In this study, we performed an interim analysis to assess the impact of implementation of the consensus management algorithm on the primary stated goals.
MATERIALS AND METHODS
Formation and Implementation of Consensus Diagnostic and Management Algorithm for Complicated Pneumonia in Children
Complicated Pneumonia Committee (CPC) members reviewed the comprehensive evidence-graded practice guidelines for pleural infections in children, published by the British Thoracic Society (BTS) in 2005.1The CPC members performed nonsystematic English-language searches in MEDLINE, using the PubMed Clinical Queries tool (key words: pneumonia, complications, abscess, necrotizing pneumonia, pleural effusion, empyema, children, treatment, and antibiotics), to identify additional more recent diagnosis and treatment studies for pediatric pneumonia. These studies were integrated into a clinical care algorithm tailored to fit the goals identified above, taking into account strength of evidence. The resulting work product was formatted as a user-friendly flowchart (Figs. 1–4 and supplementary table [Supplemental Digital Content 1, http://links.lww.com/JIM/A6]) and posted on the hospital intranet, with an accompanying evidence-based treatment guideline document, to facilitate immediate access to all caregivers. To maximize implementation and use, the algorithm was formally introduced to physicians at a dedicated educational conference in conjunction with grand rounds focusing on complicated pneumonia.
Assessment of Implementation of Algorithm
Inclusion Criteria
Children and adults aged 3 months to 20 years with primary or secondary diagnoses of empyema and/or pleural effusion in conjunction with bacterial pneumonia were identified from Children’s National Medical Center hospital databases (see description below) for inclusion in this study. Cohort 1 (n = 83) consisted of children whose conditions were diagnosed between June 28, 2007, and October 31, 2008, 16 months before implementation of the algorithm. Cohort 2 (n = 87) consisted of children whose conditions were diagnosed between November 1, 2008, and January 31, 2010, 15 months after implementation of the algorithm.
Exclusion Criteria
Because the algorithm was developed specifically for children younger than 3 months without underlying chronic or immunosuppressive medical conditions, children with diagnoses related to chronic or congenital cardiac, renal, pulmonary, rheumatologic, and immunodeficiency disorders (as determined by diagnosis codes and medical chart review) were excluded from this study.
Data Collection
The clinical and financial data system (Trendstar, McKesson, San Francisco, CA) and physician billing system (PracticePoint Manager, McKesson, San Francisco, CA) were used to collect data. Subject characteristic data including age, sex, readmission, severity level, and LOS were identified using Trendstar. Severity level was determined using the All-patient Refined Diagnosis-related Group (APR-DRG) Complexity Score, which ranges from 0 to 4, indicating minor to extreme complexity determined on the basis of the complexity levels of secondary diagnoses, in combination with the principal diagnosis and the patient’s age.2Ultrasound (U/S), computed tomography (CT), and video-assisted thoracoscopic surgery (VATS) used were identified using current procedural terminology (CPT) codes. Vancomycin use was identified by hospital charge codes. Institutional review board (IRB) approval was obtained from the institutional Human Research Protection Program to evaluate all subject information included in this study. A waiver of consent was approved by the IRB, as no subjects were prospectively enrolled in this study, no information was shared from this study, and there was no added risk to subjects by including relevant information in this study.
Statistical Analysis
A cross-sectional study analyzing the use of CT or U/S as the primary imaging modality after plain chest radiograph was undertaken. Comparison of CT and U/S use between study cohorts, as well as other outcomes (use of vancomycin, use of VATS, LOS, and readmission rate) were analyzed using the χ2 test, linear and ordinal regression, and analysis of variance. Data analysis was performed using Stata (v. 11, Stata Corp, TX) and SPSS (v. 17.0, SPSS Inc, Chicago, IL).
RESULTS
Overall Subject Characteristics
During the defined study period, 371 admissions were identified with primary or secondary diagnosis codes for empyema and/or pleural effusion in conjunction with bacterial pneumonia. After applying exclusion criteria, 170 subjects were included. Cohort 1 (preimplementation cohort) consisted of 83 subjects; cohort 2 (postimplementation cohort) consisted of 87 subjects. Mean age (cohort 1, 5.6 years; cohort 2, 5.1 yrs; P = 0.56), age distribution, sex (P = 0.30), diagnoses (P = 0.12), and severity level (P = 0.84) were similar between the 2 cohorts (Table 1). These findings suggest there was no significant difference in subject characteristics between study cohorts.
Analysis of Algorithm Implementation
There was no difference in the percentage of patients receiving supplementary diagnostic imaging after plain chest radiograph between the 2 cohorts (cohort 1, 54.2%; cohort 2, 40.2%; P = 0.07). Of patients who received supplementary imaging, there was a significant decrease in use of CT scan as the primary diagnostic imaging modality (after chest x-ray) in the postimplementation cohort (cohort 1, 60.0%; cohort 2, 17.2%; P = 0.001), accompanied by increased use of U/S as the imaging modality of choice (cohort 1, 26.7%; cohort 2, 71.4%). Readmission rate was reduced in the postimplementation cohort (cohort 1, 7.7%; cohort 2, 0%; P = 0.01). The use of VATS decreased in the postimplementation cohort (cohort 1, 44.6%; cohort 2, 28.7%; P = 0.03). These changes occurred without a concomitant increase in the use of vancomycin (P = 0.95) or hospital LOS (P = 0.4; Table 2).
In patients who received VATS drainage (n = 62), there were no significant differences between cohorts for median time from admission to VATS (cohorts 1 and 2, 2 days; P = 0.29) or median duration of chest tube drainage (cohort 1, 3 days vs cohort 2, 4 days; P = 0.10). The overall rate of pathogen identification was 44% in this group. There was a statistically nonsignificant trend for a higher rate of pathogen identification in cohort 2 (cohort 1, 33%; cohort 2, 54.1%; P = 0.12). Streptococcus pneumonia was the most commonly identified pathogen in both cohorts (27% vs 37.5%; P = 0.23).
These findings demonstrate that implementation of a comprehensive diagnosis and treatment algorithm for complicated pneumonia at our institution (which included an educational component) successfully reduced the use of CT scan for the diagnosis of pleural infections, reducing radiation exposure in pediatric patients with complicated bacterial pneumonia. In addition, we observed a decrease in VATS intervention and decrease in readmission rate. These significant changes in medical practice took place without adverse events such as increased vancomycin exposure or increased hospital LOS.
DISCUSSION
Although overall rates of bacterial pneumonia in children have declined since institution of universal pneumococcal vaccination, it has been appreciated that infectious complications of pneumonia such as empyema have continued to rise.3,4Despite this increase, a standardized treatment plan for these patients does not currently exist. The lack of concise medical literature regarding diagnostic studies, antibiotic therapy, surgical intervention, and long-term care often leads to unnecessary radiation and inappropriate antibiotic therapy. In an effort to improve management practices for children with complicated pneumonia, our institution developed a multidisciplinary consensus diagnosis and management algorithm for complicated pneumonia in children. This study evaluated several clinical outcomes associated with implementation of this algorithm for 15 months. After algorithm implementation, we demonstrated a significant reduction in unnecessary radiation exposure by effecting a shift in the diagnostic radiological modality of choice (after detection of pleural fluid on initial chest radiograph) from CT to U/S. This was accomplished without concomitant increases in adverse outcomes such as use of vancomycin or LOS, and was associated with a decrease in VATS intervention, as well as readmission rate.
Older published studies suggested that U/S could not reliably establish the stage of pleural infection.5However, more recent studies have demonstrated that chest U/S can discriminate the progressive stages of bacterial parapneumonic effusion, often obviating the need for chest CT.6Ultrasound may be more sensitive than chest radiograph in quantifying pleural effusion,7particularly in the setting of “whiteout” on chest radiograph (differentiating it from pulmonary infiltrates).8Ultrasound may help distinguish exudative pleural effusions from transudates.9In addition, U/S can estimate the size of the effusion, differentiate free from loculated pleural fluid, and determine the echogenicity of the fluid and fibrinous septations.10Ultrasound may also demonstrate pleural thickening and assist in diagnosis of effusion secondary to tuberculosis (ie, diffuse small nodules on the pleural surface),11and has been shown to distinguish fluid from solid material in the pleural space.12Although U/S may not predict failure of chest drain and fibrinolytics alone,5it can be used to guide chest drain insertion or thoracentesis,7,13–15and can assist in guided diagnostic aspiration. Other factors, including its increased availability, lack of radiation, and decreased need for sedation, has made U/S the preferred investigation in children.
Whereas CT may reveal clinically significant findings not apparent on radiography,16U/S has begun to replace CT for the reasons listed above. A study of 30 children demonstrated that CT was not helpful in differentiating empyema from parapneumonic effusion.17In a review of U/S and CT in a group of 50 adults with parapneumonic effusion requiring drainage, neither technique reliably identified the stage of the pleural effusion, although pleural thickness on CT was greater in those with frankly purulent effusions.5Computed tomography of the chest with contrast enhancement assists in delineating loculated pleural fluid and can also detect airway or parenchymal lung abnormalities, such as endobronchial obstruction or a lung abscess, as well as helping with mediastinal pathology.18,19Although unnecessary for most cases of pediatric empyema, chest CT has a role in complicated cases including initial failure to aspirate pleural fluid, patients who fail appropriate medical management, immunocompromised children, and malignancy (eg, lymphoma). Computed tomography may be useful in differentiating between complicated empyema and formation of lung abscess and may help identify impending cavitation that may not be identified on basic chest radiograph.18,20
In our study, there was no significant difference between preimplementation and postimplementation cohorts with regard to age, sex, diagnosis, or severity level—potential confounding variables that may influence imaging practices. Thus, the observed differences in imaging choice between cohorts can be attributed to implementation of the algorithm and change of practice rather than inherent differences in the cohort patient populations.
We considered that one potential effect of shifting diagnostic practices from CT to U/S was the possibility of more conservative institutional management practices, potentially increasing hospital LOS. However, this was not observed within our study cohort, suggesting that implementing an algorithm based on scientific evidence may ease physicians’ concerns that often accompany the integration of new clinical practices. Furthermore, the decrease in incidence of re-admission in the postimplementation cohort (0%) compared to the preimplementation cohort (7.7%) may in part reflect adherence to the detailed clinical inpatient monitoring outlined within the algorithm.
Analysis of nonsurgical versus surgical intervention (eg, VATS) should be considered with the involvement of the surgical and/or interventional radiology teams. Recent randomized controlled trials21,22of VATS versus fibrinolytic therapy concluded no difference in outcomes except higher cost with VATS. Other studies have shown VATS performed within 48 hours of hospitalization significantly decrease hospital LOS.4,23–25
Video-assisted thoracoscopic surgery has become recognized as a less invasive surgical option for early drainage of the pleural space6,25–28when available; it has been associated with lower hospital mortality rate, reintervention rate, LOS, time with tube thoracostomy, and time of antibiotic therapy, compared with nonoperative treatment.26,29In our study, VATS was performed less frequently after implementation of the algorithm, which suggests that implementation may have led to more judicious use of this procedure.
To evaluate the impact of algorithm implementation on limiting unnecessary antibiotic use, we evaluated the use of vancomycin. We found no significant differences in the use of vancomycin between study cohorts. However, antibiotic charge code may be a suboptimal measure of actual use, as weight-dependent antibiotic dosing in pediatric patients and inconsistent pharmacy distribution practices may lead to variability in charge codes. Ultimately, a cross-sectional or prospective study using antibiotic use as the primary outcome would be more appropriate and efficacious in determining the effect of this algorithm’s implementation on antibiotic use.
The use of billing codes, charge codes, and International Classification of Diseases, Ninth Revision codes for clinical outcomes in a cross-sectional study has inherent limitations. Outcomes such as imaging modality and diagnosis may be misrepresented, and causality between intervention (algorithm implementation) and outcomes can be difficult to directly assess. This study used historical controls; thus, other changes in the medical environment could have contributed to our observations. Therefore, we cannot be certain that any changes in practice were due solely to the implementation of the algorithm. In addition, the study was carried out at a single institution; variations in local practice and in epidemiology could lead to differences in outcomes at other institutions.
Despite these limitations, we have demonstrated that implementation of a consensus algorithm for diagnosis and management of pediatric complicated pneumonia improved patient care. Implementation successfully effected a shift from CT to U/S as the imaging modality of choice after plain film, suggesting that children with pleural infections may be diagnosed accurately using U/S, with significant reductions in radiation exposure. Moreover, implementation of the algorithm resulted in reduced re-admission rate, and a trend for increased pathogen identification, without an increase in hospital LOS or increased use of vancomycin.
Ultimately, the design and implementation of this diagnostic and management algorithm served as a critical first step in optimizing care for children with complicated bacterial pneumonia at our institution. In addition, our CPC succeeded in generating open discussion across a variety of clinical disciplines, which itself is likely to play an important role in improving health care. We plan to more formally assess the impact of algorithm implementation on our secondary goals, which include improved communication between subspecialists and hospitalists and provision of adequate postdischarge follow-up care with the appropriate care providers. Taken together, these studies will establish the framework for future prospective quality improvement and comparative efficacy studies in pediatric patients with complicated pneumonia, including future studies focused on diagnostic laboratory testing, radiographic, surgical, and antimicrobial interventions. These studies are of high priority, as they will provide the needed evidence for the optimization of our current approaches to treatment of complicated bacterial pneumonia in the pediatric population.