Hospitalization for Community-Acquired Pneumonia in Children: Effect of an Asthma Codiagnosis
BACKGROUND AND OBJECTIVE: Community-acquired pneumonia (CAP) is a common and expensive cause of hospitalization among US children, many of whom receive a codiagnosis of acute asthma. The objective of this study was to describe demographic characteristics, cost, length of stay (LOS), and adherence to clinical guidelines among these groups and to compare health care utilization and guideline adherence between them.
METHODS: This was a multicenter retrospective cohort study using data from the Pediatric Health Information System. Children aged 2 to 18 who were hospitalized with uncomplicated CAP from July 1, 2007, to June 30, 2012 were included. Demographics, LOS, total standardized cost, and clinical guideline adherence were compared between patients with CAP only and CAP plus acute asthma.
RESULTS: Among the 25 124 admissions, 57% were diagnosed with CAP only; 43% had a codiagnosis of acute asthma. The geometric mean for standardized cost was $4830; for LOS, it was 2.01 days. Eighty-four percent of patients had chest radiographs; CAP+acute asthma patients were less likely to have a blood culture performed (36% vs 62%, respectively) and more likely not to have a complete blood count performed (49% vs 27%, respectively). Greater guideline adherence was associated with higher cost at the patient-level but lower average cost per hospitalization at the hospital level. CAP+acute asthma patients had higher relative costs (11.8%) and LOS (5.6%) within hospitals and had more cost variation across hospitals, compared with patients with CAP only.
CONCLUSIONS: A codiagnosis of acute asthma is common for children with CAP. This could be from misdiagnosis or co-occurrence. Diagnostic and/or management variability appears to be greater in patients with CAP+asthma, which may increase resource utilization and LOS for these patients.
Community-acquired pneumonia (CAP) ranks among the most common and costly reasons for hospitalization of children.1 Children hospitalized with CAP often receive a diagnosis of asthma, a condition colloquially referred to as “asthmonia.” Distinguishing CAP occurring alone from CAP occurring with an asthma exacerbation can be challenging because viral respiratory tract infections predispose to bacterial pneumonia and also trigger asthma exacerbations. In addition, although CAP is typically diagnosed by chest radiograph or focal lung findings, there is evidence that atelectasis can mimic infiltrates,2 which could lead to misdiagnosis of pneumonia in a patient with asthma. Children with asthma are at increased risk for invasive pneumococcal infections,3 including pneumonia. An important challenge in the management of CAP occurring with an asthma exacerbation is that national guidelines focus treatment on a single diagnosis (CAP alone or asthma alone), and in some cases, the guidelines conflict on ideal management.4,5
The guidelines for treatment of uncomplicated CAP include obtaining bacterial cultures, viral testing, inflammatory markers, chest radiography, and the use of an aminopenicillin. Corticosteroids, a mainstay of asthma therapy, have not been proven effective for the treatment of CAP in randomized trials involving adults6 and are not recommended. There is no role for bronchodilators in the treatment of CAP. In contrast, management of an acute asthma exacerbation includes administration of oral corticosteroids and bronchodilators. Routine chest radiographs and viral testing are not recommended, and in fact, chest radiographs in pediatric asthma were chosen as one of the Choosing Wisely campaign’s procedures not to order.7
Previous studies of hospital-level adherence to CAP guidelines have shown variability; the average percent of patients receiving a chest radiograph for CAP across 29 children’s hospitals ranged from 54% to 90%, whereas the average percent of patients with a documented blood culture ranged from 0% to 92%.8 Conversely, adherence to the asthma care guidelines of using systemic corticosteroids and bronchodilators was high and consistent across hospitals.9 Although the recommendations for management of uncomplicated pneumonia alone or asthma alone are well defined, they do not provide direction for a common clinical occurrence: treating patients who are diagnosed with both. We hypothesized that this might lead to more diagnostic uncertainty, and therefore decreased adherence to clinical care guidelines, increased health care utilization, and variation in the clinical management and treatment of these patients. Understanding how patients who have pneumonia and asthma codiagnoses influence utilization might help to clarify more appropriate patient selection for guidelines use, and allow systems of care–based interventions to more accurately measure patient and cost outcomes.
We used data from a national multihospital cohort of children to determine the variation in resource utilization and adherence to the pneumonia guidelines in children with CAP with and without a concomitant diagnosis of asthma.
Data Source and Study Design
We conducted a retrospective, multicenter cohort study using data from the Pediatric Health Information System (PHIS), an administrative database that contains inpatient data from >40 pediatric hospitals in the United States. Detailed hospitalization and resource utilization data, such as demographic, diagnostic, procedural, outcome, and charge information, are contained in PHIS. Data are deidentified; however, encrypted medical record numbers permit tracking of patients within hospitals across hospitalizations. The Children’s Hospital Association and participating hospitals jointly ensure data quality as previously described.10 This study, using deidentified data, was considered exempt according to the policies of the Cincinnati Children’s Hospital Medical Center Institutional Review Board.
We created a cohort of patients with uncomplicated CAP and CAP+asthma for which the pneumonia guidelines would apply.4 We used a previously validated algorithm for identifying hospitalized patients with CAP from PHIS.11 Patients aged >3 months and <18 years between July 1, 2007, and June 30, 2012, with principal or secondary diagnosis International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes of pneumonia were eligible. We excluded patients who were unlikely to have a diagnosis of uncomplicated CAP using the following criteria: (1) receipt of antibiotics not typically used for CAP on day 0, 1, or 212; (2) no charge for antibiotics on the first hospital day; (3) an International Classification of Diseases code for viral pneumonia; (4) complex chronic condition13; (5) age <2 years (to avoid misclassification with bronchiolitis); (6) a charge for a chest computed tomography, ultrasound, or decubitus films on the first hospital day (to exclude complicated pneumonia on presentation); (7) a charge for hyperalimentation on the first hospital day; or (8) a charge for extracorporeal membrane oxygenation on the first hospital day. Three clinicians reviewed all additional diagnoses and procedures and excluded patients with diagnoses or procedures suggesting a low likelihood of CAP (eg, motor vehicle traffic accident). To better ensure patients with a typical presentation of CAP, we excluded hospitalizations for which total costs were beyond the lowest and highest fifth percentile within each hospital.1 For patients with multiple admissions meeting the inclusion criteria, we selected the first hospitalization.
The main exposure was meeting inclusion criteria for CAP and a codiagnosis of acute asthma (CAP+asthma). Patients were considered to have CAP+asthma if they had any diagnosis code of asthma and a charge for a short-acting bronchodilator use after hospital day 1. Because 35% of children admitted with a diagnosis of CAP in 1 large cohort had received corticosteroids14 and this was more common in patients in the ICU, we did not include steroid use as a specific marker for asthma to avoid confounding by severity. Patients in the CAP+asthma exposure group were compared with the clinical group who had CAP without acute asthma.
Categorized age (<5, 5–11, ≥12 years), gender, race (white, black, Asian, other); insurance (government, nongovernment); geographic region (Midwest, Northeast, South, West), mean number of hospital beds, and ICU stay (yes/no) were included in the analyses.
Measured outcomes were length of stay (LOS), total standardized cost, and guideline adherence. LOS in days was measured as a continuous variable. Total standardized cost in US dollars was calculated using a Cost Master Index (CMI) previously developed for research utilizing data from PHIS hospitals.1 Briefly, standardized costs for an entire patient hospitalization are calculated by first multiplying the billed units of each item by its standardized per unit cost and then summing these itemized costs, resulting in a total standardized hospital bill.15 All standardized costs were inflated to 2012 dollars by using the medical care services component of the Consumer Price Index. Of the 42 hospitals contained in the PHIS database, 2 were excluded from cost and LOS analyses due to missing data.
Adherence to the Pediatric Infectious Diseases Society and the Infectious Diseases Society of America clinical guidelines for the management of children with pneumonia3 was the final outcome. The following guidelines were assessed, based on availability in the PHIS dataset: receipt of chest radiograph (recommended), blood culture (recommended), respiratory viral panel testing (recommended), Mycoplasma pneumoniae testing when macrolide was ordered (recommended), complete blood count testing (not recommended), and Chlamydophila pneumoniae testing (not recommended). We considered all available recommendations regardless of strength of evidence. Adherence to guidelines on the individual level was depicted as a rate (number of guidelines followed/total guidelines measured). Adherence to guidelines at the hospital-level was depicted as the mean number of guidelines followed per patient with CAP admitted to that particular hospital.
We described variation in the total standardized costs within hospital using the geometric mean and range. Bivariable differences between the exposure groups were tested using χ2 and Wilcoxon rank-sum tests for individual-level characteristics. We used Cochran-Mantel-Haenszel statistics to test differences in demographic and clinical course characteristics between the 2 clinical groups when stratified by hospital, and log-rank tests were used for stratified tests to evaluate differences between clinical groups for continuous but skewed characteristics accounting for hospital. Analysis of variance was used to test bivariable differences for hospital-level variables (region and number of beds).
We modeled cost and LOS with a log γ model with patient-level factors as covariates and accounting for clustering of patients by hospital. These models decomposed the effects of the covariates into within-hospital and between hospital components to rule out possible confounding by hospital in these associations of interest.16,17 This decomposition permits 2 comparisons of interest: (1) the effects of the risk factor on outcome when applied to otherwise similar children within a hospital (eg, do 2 similar patients at the same hospital have different outcomes depending on guideline adherence in their management) and (2) the effects when moving from a hospital with 1 average level of a risk factor to a hospital with a different average level of the same factor (eg, do similar patients admitted to 2 hospitals with varying average guideline adherence across their patient populations have different outcomes).18 Results from this decomposed model of the within hospital comparison were then confirmed with models in which hospital was a fixed effect. We also evaluated the effect of codiagnosis on cost and LOS within hospitals by hospital, by comparing the likelihood ratios from 2 models. The first was the same model as described earlier, with hospital as a fixed effect. The second included an interaction term between hospital and clinical group (CAP or CAP+asthma). A likelihood ratio test was performed to examine cost and LOS by hospital. Data were analyzed using SAS v9.3.
Finally, we investigated the relative contribution to interhospital variance in standardized costs attributable to CAP and CAP+asthma using a mixed effects model with separate hospital-level random intercepts for the 2 groups. We compared these 2 variance components with total variance (including the residual, within-hospital component), using Stata version 13. Because log γ models for this question would not converge, we opted for identity link models for this estimation.
Patient and Hospital Characteristics
Of the 112 441 admissions eligible for study inclusion, 25 124 with CAP remained after applying exclusion criteria (Fig 1); of these, 43% had an additional diagnosis of and treatment of acute asthma.
The median age was 5.0 years (Table 1). The distribution of age group, insurance, ICU stay, and race varied widely across hospitals. When comparing the 2 clinical groups, patients hospitalized with CAP+asthma had a greater median age and were more likely to be male, to be black or “other” race/ethnicity, and to have government insurance than those with CAP only. Because patient characteristics are known to differ by hospital, we compared demographic and clinical course characteristics between the 2 clinical groups adjusting for hospital. The difference in the number of ICU days by clinical group reached statistical significance when accounting for hospital; other results were similar to the unadjusted results (Table 1). Patients with CAP+asthma were more likely to have received a macrolide antibiotic than those with CAP alone (59% vs 37%; P < .001) and to have received corticosteroids (87% vs 12%; P < .001).
Standardized cost varied between hospitals, from $3292 to $6547, with an overall geometric mean of $4830. Figure 2 displays the cost distribution by hospital and clinical group. Thirty-eight of the 40 hospitals had higher costs for the CAP+asthma patients compared with CAP alone. CAP+asthma patients had higher geometric mean cost compared with CAP alone ($5270 vs $4519) (Table 1). Multivariable analysis of total standardized cost within hospitals indicated that CAP+asthma patients were more costly compared with CAP alone (Table 2). We estimated cost variances not explained by patient-level factors between patients with CAP compared with those with CAP+asthma, and found that unexplained interhospital variation as a percentage of total variation in costs was 7.6% for CAP and 12.3% for CAP+asthma. The likelihood ratio test for within hospital cost was statistically significant, suggesting that there was some variation in total standardized cost between CAP+asthma and CAP only patients across hospitals. However, the data suggest that for the vast majority of hospitals, cost is higher among CAP+asthma patients.
The overall geometric mean for LOS was 2.01, with a range of 1.3 to 2.6 between hospitals. For ICU days, the overall mean was 1.63 with a range between hospitals from 1.0 to 2.3 (Table 1). Figure 3 shows the distribution of LOS by clinical group and hospital. Although 7 of 40 hospitals had longer LOS for CAP alone, the majority had longer LOS for CAP+asthma. Patients with CAP+asthma had longer LOS (2.1 vs 1.9); however, ICU LOS was similar (1.6 vs 1.6, P = .28). Multivariable analysis indicated that patients with CAP+asthma had a 6% longer LOS compared with CAP patients without acute asthma (Table 2). The likelihood ratio test for within hospital LOS was statistically significant, suggesting that there was some variation in LOS between CAP+asthma and CAP-only patients across hospitals. For the majority of hospitals, LOS was longer for CAP+asthma compared with CAP only patients.
When comparing the 2 clinical groups at the patient level, >80% of patients in both groups received chest radiographs (Table 1). CAP+asthma patients were less likely to have a blood culture performed (36% vs 62%) and more often did not have a CBC performed (49% vs 27%). Few patients in either group received a macrolide with M pneumoniae testing (close to 7% in each group), and few patients in either group had C pneumoniae testing. Approximately 30% in each group received respiratory viral panel testing. Across hospitals, the range of the percent of guideline adherence was wide for all of the guidelines except testing for M pneumoniae and C pneumoniae. At the patient-level, greater guideline adherence was associated with higher cost. However, contrasting adherence across hospitals, greater hospital-level guideline adherence was associated with lower average cost per hospitalization (Table 2).
We found a high percentage of cases of pneumonia that had a codiagnosis of and treatment for asthma exacerbation; although we used an algorithm for identifying CAP that has been validated with chart review and radiology data, the presence of an infiltrate on a chest x-ray may not always represent pneumonia but may in fact be atelectasis in a case of primary asthma. In addition, we may be overidentifying cases of asthma in which albuterol is being used inappropriately. Regardless of whether these cases are misdiagnoses or are truly patients with pneumonia and asthma, this study demonstrated the challenge of measuring guideline adherence in the face of codiagnoses and diagnostic uncertainty, particularly in the management of common, acute pulmonary processes in children. In this cohort of patients admitted to 42 children’s hospitals, codiagnosis resulted in broader variation in management and higher cost. These results also highlight the complexity of implementing guidelines where there is true codiagnosis or diagnostic uncertainty in carefully identifying the appropriate target population for guideline development and management so that when comparing outcomes, hospitals are using the clinical groups of patients.
Overall, CAP+asthma patients within the same hospital had higher costs and longer LOS than those with CAP alone. There was significant variability in cost and LOS among the hospitals for both CAP and CAP+asthma. Although having a second diagnosis may inherently increase overall cost, a secondary diagnosis of asthma also increased observed variation in costs across hospitals, suggesting that the addition of the second diagnosis may increase uncertainty. This could be diagnostic uncertainty, management uncertainty, or both.
Inappropriate diagnosis with a second disease process could result in overuse or misuse of nonrecommended care, and therefore higher costs. Increasing diagnostic precision and standardizing treatment of codiagnoses such as CAP+asthma may help to decrease interhospital variability in cost. In particular, adhering to the recommendation not to get a chest radiograph in children admitted with clinical asthma exacerbations might significantly decrease the overall cost and variation in care. The mean LOS for CAP+asthma of 2.1 days was longer than that reported in another PHIS study for noncomplex asthma alone (1.8 days),19 also supporting the idea that the additional diagnosis adds complexity. Future studies examining the impact of a pneumonia diagnosis on variation in care for asthma exacerbations may provide corollary information.
Overall, hospitals that had greater adherence to the CAP guidelines across their patient population had lower the overall costs for CAP and CAP+asthma, suggesting that although some guidelines may be applied inappropriately, especially for CAP+asthma, standardization of care has some association with lower costs of care. Because adherence to guidelines for both CAP and asthma are being used as quality metrics in some institutions, it may be helpful to use these as metrics for patients only when a codiagnosis is not present.
Studies using administrative data are inherently limited by the quality of the data provided by the hospitals, and there are many potential explanatory variables that are not available to us in PHIS. Although PHIS has excellent quality control standards, its data are not routinely validated with chart review. We used an algorithm for diagnosing CAP that has been validated with chart review,11 but it is nonetheless possible that there was misclassification for both the CAP and asthma diagnoses. In addition, we used antibiotic choice to select out cases of CAP, which meant that we could not include antibiotic selection in our analysis of guideline adherence. Our definition of acute asthma exacerbation may cause us to underestimate such cases. We did not include corticosteroid administration in our definition because corticosteroids, particularly at some hospitals, may be administered to children with CAP.14 This approach would bias the results to the null hypothesis. We also had a significant number of exclusions to create our final cohort; some of these may have biased our data. Although the pneumonia guidelines have recommendations related to M pneumoniae and C pneumoniae testing, the lack of timely availability of results for these tests at many institutions is likely to be an important barrier to their use. The data are retrospective, and thus we cannot assume causation. Because we limited our analyses to children >2 years of age, this may have increased the potential for asthma codiagnoses. Finally, our data may not be generalizable outside of freestanding children’s hospitals.
A codiagnosis of acute asthma is common for children with CAP, adding a layer of treatment complexity that may increase resource utilization and LOS, regardless of whether there is diagnostic misclassification or whether there are truly 2 disease processes. There is still a high degree of uncertainty surrounding the most effective and efficient way to treat CAP+asthma, and the appropriateness of existing guidelines to address the co-occurrence of these conditions. Clinicians should be clear about which diagnosis they are treating, which guideline should be applied when the clinician is unsure, or whether 2 diagnoses are clearly present. Patients with 2 diagnoses deserve to have both diagnoses treated according to evidence-based protocols. In the interim, hospitals can examine their practice patterns for CAP+asthma and if they are outliers, work to standardize and streamline care. As we increasingly rely on guidelines for the care of children hospitalized with common diseases, it is important to remember that they often have >1 diagnosis, and these guidelines should incorporate evidence about common co-occurring conditions as well as provide recommendations on their treatment.
Dr Wilson conceptualized the design and led the study and drafted the initial manuscript; Dr Torok helped with analytic planning and methods, completed statistical analyses, and helped draft the manuscript; Dr Localio led the analytic plan and advised on statistical procedures and study design, performed statistical analyses, drafted parts of the manuscript, and helped edit the manuscript; Mr Luan and Dr Mohamad completed statistical analyses, assisted with study design, and helped edit the manuscript; Drs McLeod, Srivastava, and Shah participated in the study design, analytic planning, and data quality review and edited the manuscript; and all authors approved the final manuscript as submitted.
FINANCIAL DISCLOSURE: Drs Wilson, Srivastava, and Shah were supported by the Children’s Hospital Association. The other authors have indicated that they have no financial relationships relevant to this article to disclose.
FUNDING: A grant from the Children’s Hospital Association to the Pediatric Research in Inpatient Settings Network
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Bradley JS,
- Byington CL,
- Shah SS,
- et al
- Feudtner C,
- Hays RM,
- Haynes G,
- Geyer JR,
- Neff JM,
- Koepsell TD
- Weiss AK,
- Hall M,
- Lee GE,
- Kronman MP,
- Sheffler-Collins S,
- Shah SS
- Drummond MF,
- Schulpher MJ,
- Torrance GW,
- O’Brien BJ,
- Stoddart GL
- Morse RB,
- Hall M,
- Fieldston ES,
- et al
- Copyright © 2015 by the American Academy of Pediatrics