How to cite this article: Castelán-Martínez OD, Hernández-Carbajal E, Contreras-García CE, NG Ojeda-Luna, Rivas-Ruiz R. Effectiveness of the outpatient treatment of the community-acquired pneumonia: systematic review and meta-analysis. Rev Med Inst Mex Seguro Soc. 2016;54(1):128-36.
REVIEW ARTICLES
Received: December 8th 2014
Accepted: April 28th 2015
Osvaldo Daniel Castelán-Martínez,a Elizabeth Hernández-Carbajal,b Carlos Eduardo Contreras-García,c Nancy Guadalupe Ojeda-Luna,d Rodolfo Rivas-Ruize
aFacultad de Estudios Superiores Zaragoza, Universidad Nacional Autónoma de México
bLaboratorio de Investigación en Biología del Desarrollo y Teratogénesis Experimental, Hospital Infantil de México “Federico Gómez”
cServicio de Admisión Continua, Hospital de Especialidades
dCentro de Atención Pulmonar del Sureste
eCentro de Adiestramiento en Investigación Clínica (CAIC), Coordinación de Investigación en Salud
c,eCentro Médico Nacional Siglo XXI, Instituto Mexicano del Seguro Social
a,b,c,eDistrito Federal
dMérida, Yucatán
México
Comunicación con: Rodolfo Rivas-Ruiz
Correo electrónico: rivasrodolfo@gmail.com
Background: Community-acquired pneumonia is an important cause of mortality and morbidity worldwide. Therefore, our aim was to assess the efficacy and safety of outpatient treatment of community-acquired pneumonia.
Methods: We systematically reviewed randomized clinical trials evaluating efficacy and safety of outpatient treatment (OPT) compared with inpatient treatment (IPT) of community-acquired pneumonia in patients without added co-morbidity. Relative Risk (RR) and 95 % confidence interval (95 % CI) were calculated.
Results: From 4088 reviewed articles, two articles were included for meta-analysis, including 2324 patients. One study was conducted in adults, and the other was carried out in pediatric patients. Treatment setting was not significantly associated with treatment failure (RR 0.84 [95% CI 0.68, 1.02]). Death occurred in 6 of 2324 with no difference between the two groups (RR 0.56 [95 % CI 0.12-2.61]). Finally, no differences were seen in hospital readmission between groups (RR 0.82 [95 % CI 0.52-1.30]).
Conclusion: Evidence shows that treatment setting of community-acquired pneumonia is not statistically associated with treatment failure or mortality.
Keywords: Pneumonia; Non-professional home care; Systematic review; Meta-analysis as topic.
Community-acquired pneumonia (CAP) is a serious health problem worldwide. It has been estimated that its incidence is one to five cases per 1,000 persons per year with a mortality rate of 3% and a cost of 8.4 billion dollars in the US.1 In addition, it is the sixth leading cause of death and the leading cause death by infection.2 Treatment site selection, whether outpatient or inpatient, is an important clinical decision for treating CAP because it affects the selection and route of administration of antibiotics, the intensity of medical observation, and medical resource use.3 The decision to hospitalize remains clinical. In the pediatric population it is recommended to hospitalize all patients. In adults this decision must be validated by an objective risk-assessment tool.4
There are different classification tools that have been used to categorize the severity of CAP. For example, the classification of the World Health Organization (WHO) for lower tract infections in children establishes three types: pneumonia, severe pneumonia, and very severe pneumonia; according to the criteria of the Pneumonia Severity Index (PSI), between 1 and 2 is mild, between 3 and 4 is moderate to severe, and 5 is severe); according to a pragmatic score based on the level of care needed: outpatient is mild, inpatient is moderate, and if the patient needs to be admitted to the intensive care unit, then the status is severe.5,6 It has recently been estimated that the cost of hospitalizing a patient with CAP is around 4000 euros, and that the main cost consists of nursing care.7 In addition to prevention, interventions are needed that reduce the cost of treating CAP. Scientific evidence suggests that treatment with oral amoxicillin is effective for treating children with bacteremia and clinical diagnosis of severe CAP when compared with intravenous penicillin.8 Similarly, a systematic review and meta-analysis of randomized clinical trials (called RCT) demonstrated no difference between oral and parenteral administration of antibiotics to treat CAP in children under five years.9 These findings suggest the feasibility of outpatient treatment of CAP with oral antibiotics. In our country the treatment of choice is hospitalization, especially in children under three years. In general, outpatient treatment in children has been shown to improve quality of life and reduce complications from secondary infections acquired in the hospital. With this in mind, the main objective of this systematic review and meta-analysis is to compare outpatient treatment of CAP with inpatient treatment.
This meta-analysis was conducted in accordance with the recommendations of the PRISMA statement.10
Databases and search strategy
An electronic search was made in Medline (from 1950 to July 2014), EMBASE (from 1980 to July 2014), the Cochrane Central Register of Controlled Trials (CENTRAL) (up to the second half of 2014), Lilacs-BIREME (up to July 2014), and IMBIOMED (up to July 2014). The search strategy in Medline included terms ((("Pneumonia" [Mesh]) AND (("Treatment Outcome" [Mesh]) OR ("Mortality [Mesh]) OR ("Efficacy "[textword])) AND ("Outpatients" [Mesh]) OR "Ambulatory Care" [Mesh]))). The same terms were used to search EMBASE, CENTRAL, IMBIOMED, and Lilacs. A search was also made in the reports of major conferences in pediatrics, internal medicine, and infectious diseases.
Study selection
For the main objective, RCTs were included that compared any outpatient antibiotic therapy with hospitalized patients receiving treatment for community-acquired pneumonia. Outpatient treatment could be initiated in the emergency department or as part of a strategy of early discharge (within the first 24-72 hours) in which all patients were initially treated as inpatients, and those assigned to outpatient treatment were moved to outpatient after a predefined period of time.
Two reviewers (Osvaldo Daniel Castelán-Martinez and Elizabeth Hernández-Carbajal) assessed the titles and abstracts of articles identified by the search strategy. Any potentially relevant publication was recovered in full and evaluated by the aforementioned reviewers, who also determined the final inclusion of studies in the meta-analysis. Agreement between reviewers was evaluated using statistical kappa.11
Data extraction and quality assessment
The same reviewers independently extracted data from the included trials. Data extraction was conducted using a standardized form to collect the information.
Outcome effectiveness was measured as treatment failure within thirty days after randomization. According to the literature, treatment failure was defined as one or more of the following: death, persistence, recurrence, or worsening of clinical signs or symptoms of infection, as well as any additions or modifications to the assigned intervention, including readmission. In addition, secondary measures were recorded: 1) all-cause mortality at 30 days, and 2) hospital readmission.
Synthesis and analysis of data
This meta-analysis was performed using the program Review Manager (RevMan 5.3). Analyses were done by intention to treat (ITT). The results are shown as relative risks (RR) with 95% confidence intervals (95% CI). The Mantel-Haenszel method was used for dichotomous data. The absolute risk reduction (ARR) was also calculated and the number needed to treat (NNT) was considered. Calculations were made using the calculator available at http://cmbe.net/?page_id=29612. Sensitivity analyses were made to determine whether the findings were robust and were in line with the quality and the decisions made during the review process. Sensitivity analysis was also done for sequence generation and allocation concealment. An analysis of best and worst case scenario was also done to assess the robustness of the results.
Because there was no heterogeneity between studies, a fixed effects model was used for all analyses. In addition, a statistical test for heterogeneity was performed using the Cochran Q test and the degree of heterogeneity was quantified with statistical.12-14
Figure 1 shows the flowchart of the identification and selection of studies. 4088 titles and abstracts were identified. After reviewing them, 96 were selected for evaluation in full text. In the end, only two studies met the criteria for inclusion in the meta-analysis.15,16 The reviewers had an agreement on the selection of abstracts and articles (kappa) of 0.89 [95% CI 0.68-1.00], so the search criteria were considered robust.
Figure 1 Flowchart of the identification and selection of studies
The clinical characteristics of the two included studies are shown in Table I. In both studies, patients randomized to the outpatient group were discharged immediately. No studies were found to perform early discharge. One study was done on 2100 children aged 3 to 59 months, which were reported as previously healthy,15 while the other study included adult patients;16 together, the studies included 2324 patients (ITT).
Table I Characteristics of included studies: outpatient versus inpatient | |||||||
Author/type of study | Year | Inclusion criteria | Outpatient group discharge | Outpatient | Inpatient | ||
Drug | Treatment
duration
(days) |
Drug | Treatment duration (days) |
||||
Carratalá et al. Non-blind RCT. Adults (67.5 ± 11.8). | 2005 | PSI classes II and III | Immediate | Levofloxacin oral (500 mg/day) | 10 | Levofloxacin intravenous and oral (500 mg/day) | 10 |
Hazir et al. Non-blind equivalence RCT. Children 3- 59 months | 2008 | Severe pneumonia defined by WHO | Immediate | Oral amoxicillin 80 - 90 mg/kg/day) in two doses | 5 | Initial treatment with intravenous ampicillin (100 mg/kg/day) for 48 hours, followed by 3 days of amoxicillin (80 - 90 mg/kg/day) in two doses | 5 |
PSI = Pneumonia Severity Index; WHO = World Health Organization |
The studies differed in the inclusion criteria used to classify patients. The study on children included patients with severe CAP diagnosed according to World Health Organization (WHO) criteria. The study included adult patients with low-risk CAP classified with the PSI scale.
The included studies performed treatment randomization and concealment of the procedure. Due to the nature of the intervention, neither of the studies was blind, which could introduce a selection bias in the procedure assignment. Both studies showed analysis by ITT and protocol.
The two RCTs included assessed the efficacy of CAP treatment by therapeutic failure; the outpatient group observed an incidence of treatment failure of 12.6% (147 patients out of 1162), while in the inpatient group it was 15.1% (176 patients out of 1162) with an RR of 0.84 (95% CI 0.68- 1.02) (I2 = 0%) (Figure 2).
Outpatient | Inpatient | Relative risk | ||||||
Study | Events | Total | Events | Total | Weight | Fixed effects M-H | 95% CI | |
Carratalá, 2005 | 1 | 110 | 0 | 114 | 10.9% | 3.11 | 0.13-75.49 | |
Hazir, 2008 | 1 | 1052 | 4 | 1048 | 89.1% | 0.25 | 0.03-2.22 | |
Total (CI 95%) | 1162 | 1162 | 100 | 0.56 | 0.12-2.61 | |||
Total events | 2 | 4 | ||||||
Heterogeneity: chi2 = 1.63, df = 1 (p = 0.20); L2 = 39% | ||||||||
General effect test: Z = 0.74 (p = 0.46) | ||||||||
M-H = Mantel-Haenzsel |
Figure 2 Forest plot of therapeutic failure of outpatient versus inpatient treatment. The squares to the left of the vertical line indicate a decrease in the risk of treatment failure in patients receiving outpatient treatment. The horizontal lines through the squares represent 95% confidence intervals (CI). The diamonds represent the overall risk effect (RR) of the meta-analysis and the corresponding 95% CI.
Safety was assessed by determining mortality. The group assigned to outpatient treatment had an incidence of 0.17% (two patients out of 1162), while the incidence of the group assigned to inpatient was 0.34% (four patients out of 1162) with an RR of 0.56 (CI 9% 0.12-2.61) (I2 = 39%) (Figure 3).
Outpatient | Inpatient | Relative risk | ||||||
Study | Events | Total | Events | Total | Weight | Fixed effects M-H | 95% CI | |
Carratalá, 2005 | 7 | 110 | 8 | 114 | 20.2% | 0.91 | 0.34-2.42 | |
Hazir, 2008 | 25 | 1052 | 31 | 1048 | 79.8% | 0.80 | 0.48-1.35 | |
Total (CI 95%) | 1162 | 1162 | 100 | 0.82 | 0.52-1.30 | |||
Total events | 32 | 39 | ||||||
Heterogeneity: chi2 = 0.05, df = 1 (p = 0.83;) L2 = 0% | ||||||||
General effect test: Z = 0.82 (p = 0.41) | ||||||||
M-H = Mantel-Haenzsel |
Figure 3 Forest plot of any cause of death within the first 30 days of outpatient treatment versus inpatient treatment. The squares to the left of the vertical line indicate a decrease in the risk of treatment failure in patients receiving outpatient treatment. The horizontal lines through the squares represent 95% confidence intervals (CI). The diamonds represent the overall risk effect (RR) of the meta-analysis and the corresponding 95% CI.
Finally, we explored the possibility that patients assigned to outpatient group had a higher incidence of readmission (2.75%, 32 out of 1162 patients) compared to the inpatient group (3.35%, 39 patients out of 1162). The comparison RR was 0.82 (95% CI 0.52-1.30) (Figure 4). By using forest plots no publication bias was observed in either comparison.
Outpatient | Inpatient | Relative risk | ||||||
Study | Events | Total | Events | Total | Weight | Fixed effects M-H | 95% CI | |
Carratalá, 2005 | 18 | 110 | 22 | 114 | 12.3% | 0.85 | 0.48-1.49 | |
Hazir, 2008 | 129 | 1052 | 55 ° | 1048 | 87.7% | 0.83 | 0.67-1.04 | |
Total (CI 95%) | 1162 | 1162 | 100 | 0.84 | 0.68-1.02 | |||
Total events | 147 | 176 | ||||||
Heterogeneity: chi2 = 0.00, df = 1 (p = 0.96;) L2 = 0% | ||||||||
General effect test: Z = 1.73 (p = 0.08) | ||||||||
M-H = Mantel-Haenzsel |
Figure 4 Forest plot of readmission within the first 30 days of outpatient treatment versus inpatient treatment. The squares to the left of the vertical line indicate a decrease in the risk of treatment failure in patients receiving outpatient treatment. The horizontal lines through the squares represent 95% confidence intervals (CI). The diamonds represent the overall risk effect (RR) of the meta-analysis and the corresponding 95% CI.
This systematic review aimed to assess the efficacy and safety of outpatient treatment of CAP. The meta-analysis quantitatively summarizes the evidence from RCTs comparing the efficacy of outpatient CAP treatment with inpatient treatment. The results show that the difference in efficacy and safety of outpatient treatment of CAP compared with inpatient treatment is not statistically or clinically significant. These results even show a trend favoring treatment at home (outpatient).
These results could impact on our environment, which would open the door to therapeutic alternatives for this highly prevalent disease. Mexico is a country inhabited by 10,722,147 children,17 in which CAP is the third leading cause of death. Specifically a decrease in the incidence of death associated with CAP is reported in the population age 1-4, based on which this pneumonia ranks fourth place as of 2010.18 This decrease in the mortality rate has been attributed to procedures such as vaccination against pneumococcus, since before its implementation CAP was the leading cause of death in that age group. Vaccination was introduced in our country a year before the results of the clinical trial in the pediatric population by the Pakistan health ministry. This paper reports that the introduction of the pneumococcal vaccine was added starting in 2011, after the study included in the analysis, which is from 2006,19 which impacts the possibility that the procedure could present better results than those presented in this analysis, assuming that the etiology of the disease is limited to pneumococcus, leaving viral etiologies as the most prevalent and, therefore, ambulatory treatment as the most effective.
On the other hand, it is important to consider the baseline state of the patients included in the pediatric patient RCT, because one of the weaknesses of the study is the inclusion criteria. In particular, the admission criterion takes polypnea as the only symptom, as suggested by the WHO recommendation, without adding fever or other data. This approach is useful in populations where there is no doctor. However, this single datum, while very sensitive, is not very specific for the diagnosis of CAP; that is, some patients may have other conditions, as shown by the same author in 2006 in patients under five who met the WHO criteria, but when an x-ray was assessed by several observers at the same time, this sign was associated in 14% with signs of interstitial pneumonitis and in 1.3% with lobar pneumonia.20 This may be a susceptibility bias suggesting that patients without pneumonia were included in the randomization who, therefore, had different pre-procedure probability, like patients with bronchospasm. Although the authors report that the data were adjusted, a stratified analysis would have been preferable to avoid this bias.
In our country, the clinical practice guideline for the management of CAP is more specific because it establishes that to make the diagnosis patients have to have a fever, cough, tachypnea, respiratory distress, and bronchial and alveolar crackles.21 Like this study, it does not include chest x-ray, so that the findings of this meta-analysis can be incorporated into clinical practice.
We may also assume selection bias of patients upon observing that a patient is in a moderately severe state, with oral intolerance or diarrhea, and therefore the decision is made to hospitalize them, which can change the end result by integrating more severe patients and those with a greater susceptibility to death into the inpatient group, as noted in the results.
If we evaluate the procedure employed, most clinical guidelines suggest the administration of high-dose oral amoxicillin (90 mg / kg / day) as first-line treatment for CAP. In this study, treatment with amoxicillin was highly effective with a low rate of serious adverse events. However, the etiology of pneumonia was not determined, since most cases are viral and one cannot distinguish between the effect of the antibiotic and that of the general measures carried out in the home as the cause of therapeutic success in the home, unlike the hospital.
Among the missing unspecified data we find the population characteristics. Specifically we find day care or morning care center attendance, as it is known that children who attend day care are more likely to acquire CAP. In Mexico, 25% of children go to daycare, and preschool attendance is 45%.17
Other possibly confusing variables are the use of pneumococcal vaccine and mass vaccination against influenza, epidemiological changes, and the emergence of penicillin-resistant and multidrug-resistant pneumococcus serotypes associated with carriers, especially in day cares.
The results of the meta-analysis should be studied carefully because, being dissimilar populations (adults and children) in which the etiology could be even different, the results could lead to errors in interpretation. However, the combination of the two clinical trials was made because the pathophysiology of community-acquired pneumonia is similar, because treatment at home versus in the hospital is very clear, and because the results in both groups are homogeneous, for both therapeutic failure (I2 = 0%) and death from any cause (1I2 = 39%).
Summary of main results
The meta-analysis shows that outpatient treatment of severe CAP has similar results in terms of efficacy and mortality (Table II), to inpatient treatment in adult patients and pediatric patients with mild CAP according to the pneumonia severity index, and in pediatric patients 3-59 months old with severe CAP according to WHO criteria. This happens even if we show a trend in favor of outpatient treatment.
Table II Summary of results. Outpatient versus inpatient treatment | ||||||
Result | RR | 95% CI | ARR | Percentage
95% CI |
NNT | p |
Therapeutic failure | 0.84 | 0.68-1.02 | 2.50% | NA | 41 | NS |
Mortality | 0.56 | 0.12-5.61 | 0.17% | -0.24 - 0.58% | 581 | NS |
Readmission | 0.82 | 0.52-1.30 | 0.60% | -0.80 - 2.00% | 166 | NS |
RR = risk ratio; ARR = absolute risk reduction; NNT = number needed to treat; NA = not applicable NS = not statistically significant |
Although the inclusion criteria were not standardized, since one of the studies included adults with mild CAP and the other included children with severe CAP, this did not influence the final result, as seen in the sensitivity analyses, as neither treatment is favored.
No comparison was made in terms of adverse events, since in the study with pediatric patients, no serious adverse event was reported. In the adult study, no differences were observed in adverse drug reactions between outpatient and inpatient.
Applicability of evidence
Multiple international associations2,4,22 recommend the use of risk-stratification scales to define the site of patient care. These include the pneumonia severity index (PSI), the CURB-65 scale, the American Thoracic Society criteria, and the recently-developed SMART-COP23 criteria, to determine if the patient requires hospital treatment or referral to an intensive care unit.
Using the pneumonia severity index decreases the number of hospital admissions in patients with mild CAP, with no increase in adverse outcomes.24 However, its use is complex and difficult to apply in saturated emergency departments. The CURB-65 score assigns a point to each of its variables, and is relatively easier to use and calculate, but has not been validated like the PSI, plus it requires laboratory studies, which could limit its applicability at first-level care.
Despite the existence and validity of these clinical decision-making tools, the standard management of CAP patients is hospitalization, probably due to overestimating the severity of the disease.25 This study shows that management at home does not confer increased risk of death and does confer benefits to the patient from medical, quality of life, and economic perspectives, points to consider when making the management decision.
One limitation of our study is that, due to sample size, one of the trials included did not have sufficient statistical power16 to detect effects on mortality. However, the clinical effect has no trend (according to GRADE techniques) that inpatient treatment is better than outpatient.
It must be noted that Hazir’s pediatric study included severely ill children, but not those very severely ill, that is those children found by WHO to have any of the following characteristics: inability to drink fluids, seizures, central cyanosis, Glasgow coma scale 9-10, remotely audible stridor at rest, or severe malnutrition. Out of the 6901 who were invited to the study, only 12% had signs of very severe pneumonia. These are not included in our study. This shows that outpatient treatment with high-dose amoxicillin was shown to be effective even in patients with CAP defined by WHO as severe. In this study, the multivariate model found that breastfeeding was a protective factor for treatment failure with an odds ratio (OR) of 0.65 (0.43 to 0.99), and age three to five months was found as a risk factor with an OR of 3.22 (1.87-5.52), and acute malnutrition with an OR of 1.79 (1.23-2.60).
Similar results were summarized in a meta-analysis that included only observational studies (OR 2.31, 95% CI 2.03-2.63).
Quality of evidence
The quality of included studies was rated as good. They are randomized clinical trials in which baseline was similar in both treatment arms, which made the assembly appropriate. Furthermore, no prognostic susceptibility bias was presented. We found that the procedure was not blind due to the nature of the intervention (outpatient treatment compared to home treatment); however, there was no performance bias. Over 90% of patients had complete follow-up, so we assume that there was no transfer bias and we conclude that the procedure was properly provided. The outcome (death or improvement) was evaluated the same way, and analyses were performed by both intention to treat (IT) and by protocol. For the purposes of this meta-analysis only IT results were taken into account.
Agreements and disagreements with other studies or reviews
Multiple observational studies,26-28 clinical trials,24 and even meta-analyses29 have shown that interventions in which the antibiotic is administered orally in patients with CAP are safe and effective when compared with standard treatment with intravenous antibiotics.
To date this is the first meta-analysis evaluating the efficacy and safety of outpatient treatment compared with inpatient treatment of CAP, so there are no similar studies with which to compare.
Implications for clinical practice
The evidence currently available from randomized clinical trials allows us to suggest the use of prediction scales such as the pneumonia severity index in adults to decide which CAP patients should receive outpatient or inpatient treatment, as it has been shown that treatment is equivalent in terms of efficacy and safety. In the pediatric population with severe CAP, according to WHO criteria, the use of high-dose amoxicillin is suggested, provided they do not present predictors of treatment failure.
Implications for research
It would be interesting to conduct a similar study in Mexico and determine the percentage of therapeutic failure establishing what risk factors inherent to our population should be considered in outpatient treatment. This would help determine the practical feasibility of this treatment. It is also necessary to evaluate high-risk groups where outpatient treatment is not possible, for whom a phased treatment could be implemented, and those given hospital treatment with early discharge.
Evidence shows that there are no statistically significant differences between outpatient and inpatient treatment of community-acquired pneumonia in both children and adults.
This systematic review was conducted at the Taller de Revisiones Sistemáticas y Metaanálisis of the Centro de Adiestramiento en Investigación Clínica (CAIC).
Declaración de conflicto de interés: los autores han completado y enviado la forma traducida al español de la declaración de conflictos potenciales de interés del Comité Internacional de Editores de Revistas Médicas, y no fue reportado alguno que tuviera relación con este artículo.