Matthew Test, BA

Data Source

Data for this retrospective cohort study were obtained from the Pediatric Health Information System (PHIS), which contains administrative data from 38 freestanding children's hospitals. Data quality and reliability are assured through a joint effort by the Child Health Corporation of America (Shawnee Mission, KS) and PHIS‐participating hospitals as described previously.36, 37 Encrypted medical record numbers allow for tracking of individual patients across hospitalizations. This study was reviewed and approved by the Committees for the Protection of Human Subjects at The Children's Hospital of Philadelphia (Philadelphia, PA).

Patients

Children 6‐18 years of age with CAP were eligible if they were discharged from a participating hospital between January 1, 2006 and December 31, 2008. Subjects were included if they received antibiotic therapy on the first day of hospitalization and if they satisfied one of the following International Classification of Diseases, 9th revision (ICD‐9) discharge diagnosis code criteria: 1) Principal diagnosis of M. pneumoniae pneumonia (483.0); 2) Principal diagnosis of a pneumonia‐related symptom (eg, fever, cough) (780.6 or 786.00‐786.52 [except 786.1]) and a secondary diagnosis of M. pneumoniae pneumonia; or 3) Principal diagnosis of pneumonia (481‐483.8 [except 483.0], 485‐486) and a secondary diagnosis of Mycoplasma (041.81).

Children younger than 6 years of age were excluded due to the low prevalence of M. pneumoniae infection.2, 38 Patients with comorbid conditions predisposing to severe or recurrent pneumonia (eg, cystic fibrosis, malignancy) were excluded using a previously reported classification scheme.39 In addition, we excluded patient data from 2 hospitals due to incomplete reporting of discharge information; thus data from 36 hospitals were included in this study.

Validation of Discharge Diagnosis Codes for Mycoplasma pneumoniae

To assess for misclassification of the diagnosis of M. pneumoniae, we reviewed records of a randomly selected subset of subjects from The Children's Hospital of Philadelphia; 14 of 15 patients had signs of lower respiratory tract infection in conjunction with a positive M. pneumoniae polymerase chain reaction test from nasopharyngeal washings to confirm the diagnosis of M. pneumoniae pneumonia. Hence, the positive predictive value of our algorithm for diagnosing M. pneumoniae pneumonia was 93.3%.

Study Definitions

We identified children with asthma in 2 ways. Asthma‐related hospitalizations were identified by an ICD‐9 code for asthma (493.0‐493.92) in any discharge diagnosis field during any hospitalization in the 24 months prior to the current hospitalization. Baseline controller medications were identified by receipt of inhaled corticosteroids (eg, fluticasone) or leukotriene receptor antagonists on the first day of hospitalization.

Systemic corticosteroids (either oral or intravenous) included dexamethasone, hydrocortisone, methylprednisolone, prednisolone, and prednisone. Measures of disease severity included admission to the intensive care unit within 48 hours of hospitalization, and administration of vancomycin or clindamycin, vasoactive infusions (epinephrine, norepinephrine, dopamine, and dobutamine), and invasive (endotracheal intubation) and noninvasive (continuous positive airway pressure) mechanical ventilation within 24 hours of hospitalization, as previously described.40, 41 Viral respiratory season was defined as October through March.

Measured Outcomes

The primary outcomes of interest were hospital LOS and all‐cause readmission within 28 days and 15 months after index discharge. We examined readmissions for asthma 15 months after index discharge as a secondary outcome measure because of the potential role for M. pneumoniae infection in long‐term lung dysfunction, including asthma.42 The 15‐month time frame was selected based on longitudinal data available in PHIS for the entire study cohort.

Measured Exposures

The main exposure was early initiation of macrolide therapy, defined as receipt of erythromycin, clarithromycin, or azithromycin on the first day of hospitalization.

Data Analysis

Continuous variables were described using median and interquartile range (IQR) or range values, and compared using the Wilcoxon rank‐sum test. Categorical variables were described using counts and frequencies, and compared using the chi‐square test. Multivariable linear (for LOS) and logistic (for readmission) regression analyses were performed to assess the independent association of macrolide therapy with the primary outcomes. Because the LOS data had a skewed distribution, our analyses were performed using logarithmically transformed LOS values as the dependent variable. The resulting beta‐coefficients were transformed to reflect the percent difference in LOS between subjects receiving and not receiving macrolide therapy.

Building of the multivariable models began with the inclusion of macrolide therapy. Variables associated with primary outcomes on univariate analysis (P < 0.20) were also considered for inclusion as potential confounders.43 These variables were included in the final multivariable model if they remained significant after adjusting for other factors, or if their inclusion in the model resulted in a 15% or greater change in the effect size of the primary association of interest (ie, macrolide therapy).44 Because corticosteroids also have anti‐inflammatory properties, we assessed for interactions with macrolide therapy. There was no interaction between macrolide and systemic corticosteroid therapy (P = 0.26, Likelihood ratio test), therefore our primary model adjusted for systemic corticosteroids.

Despite adjusting for systemic corticosteroid therapy in our primary analysis, residual confounding by indication for corticosteroid therapy might exist. We therefore repeated the analysis after stratifying by receipt or non‐receipt of systemic corticosteroid therapy. Because the benefit of macrolides in preventing long‐term dysfunction may be limited to those without a prior diagnosis of asthma, we repeated the analysis of readmissions within 15 months of index discharge (any readmission and asthma‐related readmissions) while limiting the cohort to those without evidence of asthma (ie, no prior asthma‐related hospitalizations and no chronic asthma medications). Because children with underlying conditions or circumstances that would predispose to prolonged hospitalizations may have been included, despite our restriction of the cohort to those without an identified chronic complex condition, we also repeated the analysis while limiting the cohort to those with a LOS 7 days. Finally, all analyses were clustered on hospital using the robust standard errors of Huber and White to account for the correlation of exposures and outcomes among children within centers.

Data were analyzed using Stata version 11 (Stata Corporation, College Station, TX). Statistical significance was determined a priori as a two‐tailed P value <0.05.

Patient Characteristics

During the study, 690 children ages 6 to 18 years met inclusion criteria. Characteristics of these patients are shown in Table 1. The median age was 10 years (IQR, 7‐13 years). Ten patients (1.4%) also had a concomitant discharge diagnosis of pneumococcal pneumonia, while 19 patients (2.7%) had a concomitant discharge diagnosis of viral pneumonia; 1 of these patients had discharge diagnoses of both viral and pneumococcal pneumonia.

Macrolide therapy was administered to 405 (58.7%) patients. Systemic corticosteroid therapy was administered to 252 (36.5%) patients. Overall, 191 (27.7%) of the 690 patients received both macrolides and systemic corticosteroids empirically, while 224 (32.5%) received neither; 61 (8.8%) received corticosteroids but not macrolides, while 214 (31.0%) received macrolides but not corticosteroids. Asthma hospitalization within the 24 months prior to admission was more common among those receiving macrolides (N = 60/405, 14.8%) than among those not receiving macrolides (N = 30/285, 10.5%) (P = 0.023). Macrolide recipients also more commonly received concomitant systemic corticosteroids (N = 191/405, 47.2%) than macrolide non‐recipients (N = 61/285, 21.4%) (P < 0.001) and more commonly received beta‐agonist therapy (N = 215/405, 53.1%) than macrolide non‐recipients (N = 113/285, 39.7%) (P = 0.001).

Length of Stay

The overall median LOS was 3 days (IQR, 2‐6 days); the median LOS was 3 days (IQR, 2‐5 days) for empiric macrolide recipients and 4 days (IQR, 2‐9 days) for non‐recipients (P < 0.001). Overall, 22.9% (N = 158) of children had an LOS 7 days and 8.8% (N = 61) of children had an LOS 14 days. The LOS was 7 days for 15.3% (N = 62) of macrolide recipients and 33.7% (N = 96) of non‐recipients. LOS was 7 days for 17.5% (N = 44) of systemic steroid recipients and 26% (N = 114) of non‐recipients. In unadjusted analysis, macrolide therapy (beta‐coefficient, 0.49; 95% confidence interval [CI]: 0.72 to 0.25; P < 0.001) and systemic corticosteroid administration (beta‐coefficient, 0.26; CI: 0.37 to 0.14; P < 0.001) were associated with shorter hospital LOS (Appendix 1).

In multivariable analysis, macrolide therapy remained associated with a shorter LOS (Table 2; Appendix 2). Systemic corticosteroid administration was associated with a 23% shorter LOS (adjusted beta‐coefficient, 0.26; 95% CI: 0.39 to 0.14; P < 0.001). In contrast, previous hospitalization for asthma was associated with a 31% longer LOS (adjusted beta‐coefficient, 0.27; 95% CI: 0.09‐0.045; P = 0.004). Receipt of beta‐agonist therapy or chronic asthma medications were not associated with significant differences in LOS. In analysis stratified by receipt or non‐receipt of concomitant systemic corticosteroid therapy, empiric macrolide therapy remained associated with a significantly shorter LOS in both systemic corticosteroid recipients and non‐recipient (Table 4). When the cohort was restricted to subjects with a LOS 7 days, macrolide therapy remained significantly associated with a shorter LOS (adjusted percent change, 20%; 95% CI: 32% to 5%; P = 0.015).

Readmission

Overall, 8 children (1.2%) were readmitted for pneumonia‐associated conditions within 28 days of index discharge. Readmission occurred in 1.2% of macrolide recipients and 1.1% of non‐recipients (P = 0.83) (Table 4). In unadjusted analysis, neither macrolide therapy (odds ratio [OR], 1.18; 95% CI: 0.25‐5.45; P = 0.84) nor systemic corticosteroid administration (OR, 1.04; 95% CI: 0.27‐4.10; P = 0.95) was associated with 28‐day readmission (Appendix 3). In multivariable analysis, empiric macrolide therapy was not associated with 28‐day readmission in the overall cohort (Table 2; Appendix 4)), or when the analysis was stratified by receipt or non‐receipt of concomitant systemic corticosteroid therapy (Table 3).

Overall, 160 children (23.2%) were readmitted within 15 months of index discharge; 95 were readmitted for asthma during this time (Table 3). Overall readmission occurred in 23.7% of macrolide recipients and 22.5% of macrolide non‐recipients (P = 0.702). Asthma readmission occurred in 15.1% of macrolide recipients and 11.9% of macrolide non‐recipients (P = 0.240). In unadjusted analysis, empiric macrolide therapy was not significantly associated with any readmission within 15 months (OR, 1.07; 95% CI: 0.69‐1.68; P = 0.759) or with asthma‐related readmission within 15 months (OR, 1.31; 95% CI: 0.73‐ 2.36; P = 0.369). In multivariable analysis, neither any readmission nor asthma readmission within 15 months was associated with empiric macrolide therapy overall (Table 2) or when stratified by receipt or non‐receipt of concomitant systemic corticosteroid therapy (Table 3).

The analyses for readmissions within 15 months of index discharge were repeated while limiting the cohort to those without prior asthma hospitalizations or chronic asthma medications. In this subset of patients, readmissions for any reason occurred in 55 (18.6%) of 295 macrolide recipients and 50 (22.0%) of 227 non‐recipients. The difference was not statistically significant in multivariable analysis (adjusted odds ratio, 0.79; 95% CI: 0.41‐1.51; P = 0.47). Readmissions for asthma occurred in 30 (10.2%) of 295 macrolide recipients and 26 (11.5%) of 227 non‐recipients; this difference was also not significant in multivariable analysis (adjusted odds ratio, 0.83; 95% CI: 0.36‐1.93; P = 0.83). The magnitude of the estimate of effect for 28‐day and 15‐month readmissions, and 15‐month asthma hospitalizations, was similar to the primary analysis when the cohort was restricted to subjects with a LOS 7 days.

Data Source

Data for this retrospective cohort study were obtained from the Pediatric Health Information System (PHIS), which contains administrative data from 38 freestanding children's hospitals. Data quality and reliability are assured through a joint effort by the Child Health Corporation of America (Shawnee Mission, KS) and PHIS‐participating hospitals as described previously.36, 37 Encrypted medical record numbers allow for tracking of individual patients across hospitalizations. This study was reviewed and approved by the Committees for the Protection of Human Subjects at The Children's Hospital of Philadelphia (Philadelphia, PA).

Patients

Children 6‐18 years of age with CAP were eligible if they were discharged from a participating hospital between January 1, 2006 and December 31, 2008. Subjects were included if they received antibiotic therapy on the first day of hospitalization and if they satisfied one of the following International Classification of Diseases, 9th revision (ICD‐9) discharge diagnosis code criteria: 1) Principal diagnosis of M. pneumoniae pneumonia (483.0); 2) Principal diagnosis of a pneumonia‐related symptom (eg, fever, cough) (780.6 or 786.00‐786.52 [except 786.1]) and a secondary diagnosis of M. pneumoniae pneumonia; or 3) Principal diagnosis of pneumonia (481‐483.8 [except 483.0], 485‐486) and a secondary diagnosis of Mycoplasma (041.81).

Children younger than 6 years of age were excluded due to the low prevalence of M. pneumoniae infection.2, 38 Patients with comorbid conditions predisposing to severe or recurrent pneumonia (eg, cystic fibrosis, malignancy) were excluded using a previously reported classification scheme.39 In addition, we excluded patient data from 2 hospitals due to incomplete reporting of discharge information; thus data from 36 hospitals were included in this study.

Validation of Discharge Diagnosis Codes for Mycoplasma pneumoniae

To assess for misclassification of the diagnosis of M. pneumoniae, we reviewed records of a randomly selected subset of subjects from The Children's Hospital of Philadelphia; 14 of 15 patients had signs of lower respiratory tract infection in conjunction with a positive M. pneumoniae polymerase chain reaction test from nasopharyngeal washings to confirm the diagnosis of M. pneumoniae pneumonia. Hence, the positive predictive value of our algorithm for diagnosing M. pneumoniae pneumonia was 93.3%.

Study Definitions

We identified children with asthma in 2 ways. Asthma‐related hospitalizations were identified by an ICD‐9 code for asthma (493.0‐493.92) in any discharge diagnosis field during any hospitalization in the 24 months prior to the current hospitalization. Baseline controller medications were identified by receipt of inhaled corticosteroids (eg, fluticasone) or leukotriene receptor antagonists on the first day of hospitalization.

Systemic corticosteroids (either oral or intravenous) included dexamethasone, hydrocortisone, methylprednisolone, prednisolone, and prednisone. Measures of disease severity included admission to the intensive care unit within 48 hours of hospitalization, and administration of vancomycin or clindamycin, vasoactive infusions (epinephrine, norepinephrine, dopamine, and dobutamine), and invasive (endotracheal intubation) and noninvasive (continuous positive airway pressure) mechanical ventilation within 24 hours of hospitalization, as previously described.40, 41 Viral respiratory season was defined as October through March.

Measured Outcomes

The primary outcomes of interest were hospital LOS and all‐cause readmission within 28 days and 15 months after index discharge. We examined readmissions for asthma 15 months after index discharge as a secondary outcome measure because of the potential role for M. pneumoniae infection in long‐term lung dysfunction, including asthma.42 The 15‐month time frame was selected based on longitudinal data available in PHIS for the entire study cohort.

Measured Exposures

The main exposure was early initiation of macrolide therapy, defined as receipt of erythromycin, clarithromycin, or azithromycin on the first day of hospitalization.

Data Analysis

Continuous variables were described using median and interquartile range (IQR) or range values, and compared using the Wilcoxon rank‐sum test. Categorical variables were described using counts and frequencies, and compared using the chi‐square test. Multivariable linear (for LOS) and logistic (for readmission) regression analyses were performed to assess the independent association of macrolide therapy with the primary outcomes. Because the LOS data had a skewed distribution, our analyses were performed using logarithmically transformed LOS values as the dependent variable. The resulting beta‐coefficients were transformed to reflect the percent difference in LOS between subjects receiving and not receiving macrolide therapy.

Building of the multivariable models began with the inclusion of macrolide therapy. Variables associated with primary outcomes on univariate analysis (P < 0.20) were also considered for inclusion as potential confounders.43 These variables were included in the final multivariable model if they remained significant after adjusting for other factors, or if their inclusion in the model resulted in a 15% or greater change in the effect size of the primary association of interest (ie, macrolide therapy).44 Because corticosteroids also have anti‐inflammatory properties, we assessed for interactions with macrolide therapy. There was no interaction between macrolide and systemic corticosteroid therapy (P = 0.26, Likelihood ratio test), therefore our primary model adjusted for systemic corticosteroids.

Despite adjusting for systemic corticosteroid therapy in our primary analysis, residual confounding by indication for corticosteroid therapy might exist. We therefore repeated the analysis after stratifying by receipt or non‐receipt of systemic corticosteroid therapy. Because the benefit of macrolides in preventing long‐term dysfunction may be limited to those without a prior diagnosis of asthma, we repeated the analysis of readmissions within 15 months of index discharge (any readmission and asthma‐related readmissions) while limiting the cohort to those without evidence of asthma (ie, no prior asthma‐related hospitalizations and no chronic asthma medications). Because children with underlying conditions or circumstances that would predispose to prolonged hospitalizations may have been included, despite our restriction of the cohort to those without an identified chronic complex condition, we also repeated the analysis while limiting the cohort to those with a LOS 7 days. Finally, all analyses were clustered on hospital using the robust standard errors of Huber and White to account for the correlation of exposures and outcomes among children within centers.

Data were analyzed using Stata version 11 (Stata Corporation, College Station, TX). Statistical significance was determined a priori as a two‐tailed P value <0.05.

Patient Characteristics

During the study, 690 children ages 6 to 18 years met inclusion criteria. Characteristics of these patients are shown in Table 1. The median age was 10 years (IQR, 7‐13 years). Ten patients (1.4%) also had a concomitant discharge diagnosis of pneumococcal pneumonia, while 19 patients (2.7%) had a concomitant discharge diagnosis of viral pneumonia; 1 of these patients had discharge diagnoses of both viral and pneumococcal pneumonia.

Macrolide therapy was administered to 405 (58.7%) patients. Systemic corticosteroid therapy was administered to 252 (36.5%) patients. Overall, 191 (27.7%) of the 690 patients received both macrolides and systemic corticosteroids empirically, while 224 (32.5%) received neither; 61 (8.8%) received corticosteroids but not macrolides, while 214 (31.0%) received macrolides but not corticosteroids. Asthma hospitalization within the 24 months prior to admission was more common among those receiving macrolides (N = 60/405, 14.8%) than among those not receiving macrolides (N = 30/285, 10.5%) (P = 0.023). Macrolide recipients also more commonly received concomitant systemic corticosteroids (N = 191/405, 47.2%) than macrolide non‐recipients (N = 61/285, 21.4%) (P < 0.001) and more commonly received beta‐agonist therapy (N = 215/405, 53.1%) than macrolide non‐recipients (N = 113/285, 39.7%) (P = 0.001).

Length of Stay

The overall median LOS was 3 days (IQR, 2‐6 days); the median LOS was 3 days (IQR, 2‐5 days) for empiric macrolide recipients and 4 days (IQR, 2‐9 days) for non‐recipients (P < 0.001). Overall, 22.9% (N = 158) of children had an LOS 7 days and 8.8% (N = 61) of children had an LOS 14 days. The LOS was 7 days for 15.3% (N = 62) of macrolide recipients and 33.7% (N = 96) of non‐recipients. LOS was 7 days for 17.5% (N = 44) of systemic steroid recipients and 26% (N = 114) of non‐recipients. In unadjusted analysis, macrolide therapy (beta‐coefficient, 0.49; 95% confidence interval [CI]: 0.72 to 0.25; P < 0.001) and systemic corticosteroid administration (beta‐coefficient, 0.26; CI: 0.37 to 0.14; P < 0.001) were associated with shorter hospital LOS (Appendix 1).

In multivariable analysis, macrolide therapy remained associated with a shorter LOS (Table 2; Appendix 2). Systemic corticosteroid administration was associated with a 23% shorter LOS (adjusted beta‐coefficient, 0.26; 95% CI: 0.39 to 0.14; P < 0.001). In contrast, previous hospitalization for asthma was associated with a 31% longer LOS (adjusted beta‐coefficient, 0.27; 95% CI: 0.09‐0.045; P = 0.004). Receipt of beta‐agonist therapy or chronic asthma medications were not associated with significant differences in LOS. In analysis stratified by receipt or non‐receipt of concomitant systemic corticosteroid therapy, empiric macrolide therapy remained associated with a significantly shorter LOS in both systemic corticosteroid recipients and non‐recipient (Table 4). When the cohort was restricted to subjects with a LOS 7 days, macrolide therapy remained significantly associated with a shorter LOS (adjusted percent change, 20%; 95% CI: 32% to 5%; P = 0.015).

Readmission

Overall, 8 children (1.2%) were readmitted for pneumonia‐associated conditions within 28 days of index discharge. Readmission occurred in 1.2% of macrolide recipients and 1.1% of non‐recipients (P = 0.83) (Table 4). In unadjusted analysis, neither macrolide therapy (odds ratio [OR], 1.18; 95% CI: 0.25‐5.45; P = 0.84) nor systemic corticosteroid administration (OR, 1.04; 95% CI: 0.27‐4.10; P = 0.95) was associated with 28‐day readmission (Appendix 3). In multivariable analysis, empiric macrolide therapy was not associated with 28‐day readmission in the overall cohort (Table 2; Appendix 4)), or when the analysis was stratified by receipt or non‐receipt of concomitant systemic corticosteroid therapy (Table 3).

Overall, 160 children (23.2%) were readmitted within 15 months of index discharge; 95 were readmitted for asthma during this time (Table 3). Overall readmission occurred in 23.7% of macrolide recipients and 22.5% of macrolide non‐recipients (P = 0.702). Asthma readmission occurred in 15.1% of macrolide recipients and 11.9% of macrolide non‐recipients (P = 0.240). In unadjusted analysis, empiric macrolide therapy was not significantly associated with any readmission within 15 months (OR, 1.07; 95% CI: 0.69‐1.68; P = 0.759) or with asthma‐related readmission within 15 months (OR, 1.31; 95% CI: 0.73‐ 2.36; P = 0.369). In multivariable analysis, neither any readmission nor asthma readmission within 15 months was associated with empiric macrolide therapy overall (Table 2) or when stratified by receipt or non‐receipt of concomitant systemic corticosteroid therapy (Table 3).

The analyses for readmissions within 15 months of index discharge were repeated while limiting the cohort to those without prior asthma hospitalizations or chronic asthma medications. In this subset of patients, readmissions for any reason occurred in 55 (18.6%) of 295 macrolide recipients and 50 (22.0%) of 227 non‐recipients. The difference was not statistically significant in multivariable analysis (adjusted odds ratio, 0.79; 95% CI: 0.41‐1.51; P = 0.47). Readmissions for asthma occurred in 30 (10.2%) of 295 macrolide recipients and 26 (11.5%) of 227 non‐recipients; this difference was also not significant in multivariable analysis (adjusted odds ratio, 0.83; 95% CI: 0.36‐1.93; P = 0.83). The magnitude of the estimate of effect for 28‐day and 15‐month readmissions, and 15‐month asthma hospitalizations, was similar to the primary analysis when the cohort was restricted to subjects with a LOS 7 days.

Data Source

Data for this retrospective cohort study were obtained from the Pediatric Health Information System (PHIS), which contains administrative data from 38 freestanding children's hospitals. Data quality and reliability are assured through a joint effort by the Child Health Corporation of America (Shawnee Mission, KS) and PHIS‐participating hospitals as described previously.36, 37 Encrypted medical record numbers allow for tracking of individual patients across hospitalizations. This study was reviewed and approved by the Committees for the Protection of Human Subjects at The Children's Hospital of Philadelphia (Philadelphia, PA).

Patients

Children 6‐18 years of age with CAP were eligible if they were discharged from a participating hospital between January 1, 2006 and December 31, 2008. Subjects were included if they received antibiotic therapy on the first day of hospitalization and if they satisfied one of the following International Classification of Diseases, 9th revision (ICD‐9) discharge diagnosis code criteria: 1) Principal diagnosis of M. pneumoniae pneumonia (483.0); 2) Principal diagnosis of a pneumonia‐related symptom (eg, fever, cough) (780.6 or 786.00‐786.52 [except 786.1]) and a secondary diagnosis of M. pneumoniae pneumonia; or 3) Principal diagnosis of pneumonia (481‐483.8 [except 483.0], 485‐486) and a secondary diagnosis of Mycoplasma (041.81).

Children younger than 6 years of age were excluded due to the low prevalence of M. pneumoniae infection.2, 38 Patients with comorbid conditions predisposing to severe or recurrent pneumonia (eg, cystic fibrosis, malignancy) were excluded using a previously reported classification scheme.39 In addition, we excluded patient data from 2 hospitals due to incomplete reporting of discharge information; thus data from 36 hospitals were included in this study.

Validation of Discharge Diagnosis Codes for Mycoplasma pneumoniae

To assess for misclassification of the diagnosis of M. pneumoniae, we reviewed records of a randomly selected subset of subjects from The Children's Hospital of Philadelphia; 14 of 15 patients had signs of lower respiratory tract infection in conjunction with a positive M. pneumoniae polymerase chain reaction test from nasopharyngeal washings to confirm the diagnosis of M. pneumoniae pneumonia. Hence, the positive predictive value of our algorithm for diagnosing M. pneumoniae pneumonia was 93.3%.

Study Definitions

We identified children with asthma in 2 ways. Asthma‐related hospitalizations were identified by an ICD‐9 code for asthma (493.0‐493.92) in any discharge diagnosis field during any hospitalization in the 24 months prior to the current hospitalization. Baseline controller medications were identified by receipt of inhaled corticosteroids (eg, fluticasone) or leukotriene receptor antagonists on the first day of hospitalization.

Systemic corticosteroids (either oral or intravenous) included dexamethasone, hydrocortisone, methylprednisolone, prednisolone, and prednisone. Measures of disease severity included admission to the intensive care unit within 48 hours of hospitalization, and administration of vancomycin or clindamycin, vasoactive infusions (epinephrine, norepinephrine, dopamine, and dobutamine), and invasive (endotracheal intubation) and noninvasive (continuous positive airway pressure) mechanical ventilation within 24 hours of hospitalization, as previously described.40, 41 Viral respiratory season was defined as October through March.

Measured Outcomes

The primary outcomes of interest were hospital LOS and all‐cause readmission within 28 days and 15 months after index discharge. We examined readmissions for asthma 15 months after index discharge as a secondary outcome measure because of the potential role for M. pneumoniae infection in long‐term lung dysfunction, including asthma.42 The 15‐month time frame was selected based on longitudinal data available in PHIS for the entire study cohort.

Measured Exposures

The main exposure was early initiation of macrolide therapy, defined as receipt of erythromycin, clarithromycin, or azithromycin on the first day of hospitalization.

Data Analysis

Continuous variables were described using median and interquartile range (IQR) or range values, and compared using the Wilcoxon rank‐sum test. Categorical variables were described using counts and frequencies, and compared using the chi‐square test. Multivariable linear (for LOS) and logistic (for readmission) regression analyses were performed to assess the independent association of macrolide therapy with the primary outcomes. Because the LOS data had a skewed distribution, our analyses were performed using logarithmically transformed LOS values as the dependent variable. The resulting beta‐coefficients were transformed to reflect the percent difference in LOS between subjects receiving and not receiving macrolide therapy.

Building of the multivariable models began with the inclusion of macrolide therapy. Variables associated with primary outcomes on univariate analysis (P < 0.20) were also considered for inclusion as potential confounders.43 These variables were included in the final multivariable model if they remained significant after adjusting for other factors, or if their inclusion in the model resulted in a 15% or greater change in the effect size of the primary association of interest (ie, macrolide therapy).44 Because corticosteroids also have anti‐inflammatory properties, we assessed for interactions with macrolide therapy. There was no interaction between macrolide and systemic corticosteroid therapy (P = 0.26, Likelihood ratio test), therefore our primary model adjusted for systemic corticosteroids.

Despite adjusting for systemic corticosteroid therapy in our primary analysis, residual confounding by indication for corticosteroid therapy might exist. We therefore repeated the analysis after stratifying by receipt or non‐receipt of systemic corticosteroid therapy. Because the benefit of macrolides in preventing long‐term dysfunction may be limited to those without a prior diagnosis of asthma, we repeated the analysis of readmissions within 15 months of index discharge (any readmission and asthma‐related readmissions) while limiting the cohort to those without evidence of asthma (ie, no prior asthma‐related hospitalizations and no chronic asthma medications). Because children with underlying conditions or circumstances that would predispose to prolonged hospitalizations may have been included, despite our restriction of the cohort to those without an identified chronic complex condition, we also repeated the analysis while limiting the cohort to those with a LOS 7 days. Finally, all analyses were clustered on hospital using the robust standard errors of Huber and White to account for the correlation of exposures and outcomes among children within centers.

Data were analyzed using Stata version 11 (Stata Corporation, College Station, TX). Statistical significance was determined a priori as a two‐tailed P value <0.05.

Patient Characteristics

During the study, 690 children ages 6 to 18 years met inclusion criteria. Characteristics of these patients are shown in Table 1. The median age was 10 years (IQR, 7‐13 years). Ten patients (1.4%) also had a concomitant discharge diagnosis of pneumococcal pneumonia, while 19 patients (2.7%) had a concomitant discharge diagnosis of viral pneumonia; 1 of these patients had discharge diagnoses of both viral and pneumococcal pneumonia.

Macrolide therapy was administered to 405 (58.7%) patients. Systemic corticosteroid therapy was administered to 252 (36.5%) patients. Overall, 191 (27.7%) of the 690 patients received both macrolides and systemic corticosteroids empirically, while 224 (32.5%) received neither; 61 (8.8%) received corticosteroids but not macrolides, while 214 (31.0%) received macrolides but not corticosteroids. Asthma hospitalization within the 24 months prior to admission was more common among those receiving macrolides (N = 60/405, 14.8%) than among those not receiving macrolides (N = 30/285, 10.5%) (P = 0.023). Macrolide recipients also more commonly received concomitant systemic corticosteroids (N = 191/405, 47.2%) than macrolide non‐recipients (N = 61/285, 21.4%) (P < 0.001) and more commonly received beta‐agonist therapy (N = 215/405, 53.1%) than macrolide non‐recipients (N = 113/285, 39.7%) (P = 0.001).

Length of Stay

The overall median LOS was 3 days (IQR, 2‐6 days); the median LOS was 3 days (IQR, 2‐5 days) for empiric macrolide recipients and 4 days (IQR, 2‐9 days) for non‐recipients (P < 0.001). Overall, 22.9% (N = 158) of children had an LOS 7 days and 8.8% (N = 61) of children had an LOS 14 days. The LOS was 7 days for 15.3% (N = 62) of macrolide recipients and 33.7% (N = 96) of non‐recipients. LOS was 7 days for 17.5% (N = 44) of systemic steroid recipients and 26% (N = 114) of non‐recipients. In unadjusted analysis, macrolide therapy (beta‐coefficient, 0.49; 95% confidence interval [CI]: 0.72 to 0.25; P < 0.001) and systemic corticosteroid administration (beta‐coefficient, 0.26; CI: 0.37 to 0.14; P < 0.001) were associated with shorter hospital LOS (Appendix 1).

In multivariable analysis, macrolide therapy remained associated with a shorter LOS (Table 2; Appendix 2). Systemic corticosteroid administration was associated with a 23% shorter LOS (adjusted beta‐coefficient, 0.26; 95% CI: 0.39 to 0.14; P < 0.001). In contrast, previous hospitalization for asthma was associated with a 31% longer LOS (adjusted beta‐coefficient, 0.27; 95% CI: 0.09‐0.045; P = 0.004). Receipt of beta‐agonist therapy or chronic asthma medications were not associated with significant differences in LOS. In analysis stratified by receipt or non‐receipt of concomitant systemic corticosteroid therapy, empiric macrolide therapy remained associated with a significantly shorter LOS in both systemic corticosteroid recipients and non‐recipient (Table 4). When the cohort was restricted to subjects with a LOS 7 days, macrolide therapy remained significantly associated with a shorter LOS (adjusted percent change, 20%; 95% CI: 32% to 5%; P = 0.015).

Readmission

Overall, 8 children (1.2%) were readmitted for pneumonia‐associated conditions within 28 days of index discharge. Readmission occurred in 1.2% of macrolide recipients and 1.1% of non‐recipients (P = 0.83) (Table 4). In unadjusted analysis, neither macrolide therapy (odds ratio [OR], 1.18; 95% CI: 0.25‐5.45; P = 0.84) nor systemic corticosteroid administration (OR, 1.04; 95% CI: 0.27‐4.10; P = 0.95) was associated with 28‐day readmission (Appendix 3). In multivariable analysis, empiric macrolide therapy was not associated with 28‐day readmission in the overall cohort (Table 2; Appendix 4)), or when the analysis was stratified by receipt or non‐receipt of concomitant systemic corticosteroid therapy (Table 3).

Overall, 160 children (23.2%) were readmitted within 15 months of index discharge; 95 were readmitted for asthma during this time (Table 3). Overall readmission occurred in 23.7% of macrolide recipients and 22.5% of macrolide non‐recipients (P = 0.702). Asthma readmission occurred in 15.1% of macrolide recipients and 11.9% of macrolide non‐recipients (P = 0.240). In unadjusted analysis, empiric macrolide therapy was not significantly associated with any readmission within 15 months (OR, 1.07; 95% CI: 0.69‐1.68; P = 0.759) or with asthma‐related readmission within 15 months (OR, 1.31; 95% CI: 0.73‐ 2.36; P = 0.369). In multivariable analysis, neither any readmission nor asthma readmission within 15 months was associated with empiric macrolide therapy overall (Table 2) or when stratified by receipt or non‐receipt of concomitant systemic corticosteroid therapy (Table 3).

The analyses for readmissions within 15 months of index discharge were repeated while limiting the cohort to those without prior asthma hospitalizations or chronic asthma medications. In this subset of patients, readmissions for any reason occurred in 55 (18.6%) of 295 macrolide recipients and 50 (22.0%) of 227 non‐recipients. The difference was not statistically significant in multivariable analysis (adjusted odds ratio, 0.79; 95% CI: 0.41‐1.51; P = 0.47). Readmissions for asthma occurred in 30 (10.2%) of 295 macrolide recipients and 26 (11.5%) of 227 non‐recipients; this difference was also not significant in multivariable analysis (adjusted odds ratio, 0.83; 95% CI: 0.36‐1.93; P = 0.83). The magnitude of the estimate of effect for 28‐day and 15‐month readmissions, and 15‐month asthma hospitalizations, was similar to the primary analysis when the cohort was restricted to subjects with a LOS 7 days.