A Systematic Review and Meta-analysis of Ventilator-associated Pneumonia in Adults in Asia: An Analysis of National Income Level on Incidence and Etiology

Abstract Background Ventilator-associated pneumonia (VAP) is the commonest hospital-acquired infection (HAI) in intensive care. In Asia, VAP is increasingly caused by resistant gram-negative organisms. Despite the global antimicrobial resistance crisis, the epidemiology of VAP is poorly documented in Asia. Methods We systematically reviewed literature published on Ovid Medline, Embase Classic, and Embase from 1 January 1990 to 17 August 2017 to estimate incidence, prevalence, and etiology of VAP. We performed a meta-analysis to give pooled rates and rates by country income level. Results Pooled incidence density of VAP was high in lower- and upper-middle-income countries and lower in high-income countries (18.5, 15.2, and 9.0 per 1000 ventilator-days, respectively). Acinetobacter baumannii (n = 3687 [26%]) and Pseudomonas aeruginosa (n = 3176 [22%]) were leading causes of VAP; Staphylococcus aureus caused 14% (n = 1999). Carbapenem resistance was common (57.1%). Conclusions VAP remains a common cause of HAI, especially in low- and middle-income countries, and antibiotic resistance is high.

Pakistan, Qatar, Saudi Arabia, Singapore, South Korea, Sri Lanka, Syria, Taiwan, Tajikistan, Thailand, The Philippines, Turkmenistan, Uzbekistan, United Arab Emirates, Vietnam, Yemen. Articles were included for full screening if the abstract indicated there was prospective data on etiology, incidence density, period prevalence, or any details on the economic cost of VAP (see Figure 1 for details regarding inclusion of studies). Two independent reviewers assessed the titles and abstracts of papers for inclusion and any discrepancies were referred to a third independent reviewer. References of papers included were also assessed and included if they fit the inclusion criteria. Data were extracted onto a standardized form that was trialed on the first 3 papers and implemented after minor adjustments. Studies were categorized as those that gave period prevalence data (also referred to as cumulative incidence, defined as the number of episodes of VAP per 100 patients over a period of time, usually their ICU admission), incidence density data (number of episodes of VAP per 1000 ventilator-days, included if the total number of ventilator-days was also given), microbiology data with or without susceptibility, and any data on costing of VAP. Microbiology data were only included when the total number of organisms was given. Antibiotic resistance was classified using the European Centre for Disease Prevention and Control guidelines on surveillance of HAIs in the ICU [15]. Cost of VAP was collected in US dollars where possible and converted from local currency at the average rate for the year data was collected in the study. Study quality was assessed using the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist for observational cohort and case-control studies, and the Consolidated Standards of Reporting Trials (CONSORT) checklist was used for randomized trials. Grading was determined by percentage completion of the checklist; high, ≥80% of items; moderate, 50%-80%; poor, <50% [16].
Countries were defined according to World Bank classification of income level as low income (LIC), lower-middle income (LMIC), upper-middle income (UMIC), and high income (HIC) at the time the study was published [17]. There were only 2 studies from LICs (both from Nepal) and therefore these were included with LMICs for all group analysis. Countries were defined as tropical or nontropical based on their geographical latitude.
Data were analyzed using Stata software version 14 and Comprehensive Meta-analysis Software (www.meta-analysis.com). Data were pooled in all studies that reported the same outcome measures. The outcome measures of interest were incidence density as rate of VAP per 1000 ventilator-days with total number of ventilator-days as a denominator; period prevalence as rate of VAP per 100 patients over a given period of time; total number of microorganisms identified per country and proportion of individual species; and antibiotic resistance rates for a species as percentage of resistant organisms of that species by total number of organisms of that species identified per country. Meta-analysis was conducted on incidence density and period prevalence by weighted average estimates from individual studies. Forest plots were generated by Stata and meta-regression was used to analyze comparable groupings-country, income level, year published, study quality, study method (randomized controlled trials, observational cohort, pre-/postintervention, quasi-experimental), hospital type (specialist, tertiary/university, district, multiple), and ICU type (general, medical, surgical, neurological, trauma, coronary, multiple). Studies were assessed for heterogeneity using I 2 statistic and χ 2 test (taking P < .05 to indicate statistically significant heterogeneity). When significant heterogeneity was found, then random-effects meta-analysis was used [18]. A funnel plot was used to determine the probability of publication bias. Comparisons of rates were analyzed using mixed effect meta-regression, an unrestricted maximum likelihood model.
Microbiological identification and susceptibility data were individually pooled by country and summary measures presented without use of meta-analysis.

RESULTS
A total of 88 studies covering 22 countries were included in the final data extraction. One hundred sixty-four were identified for full article review, and 76 were excluded for the following reasons: non-English language (n = 10), pediatric (n = 5), lacking specific data (n = 21), retrospective (n = 7), non-journal article (n = 31), and unable to retrieve (n = 2) ( Figure 1). The majority of studies were published in the last 10 years (2008 onward, 67/88 [76%]). There were 69 (78%) studies of high quality, 19 (22%) of moderate quality, and none of poor quality. Data on VAP incidence density were available from 31 articles and period prevalence from 74 articles. Microbiology data were available from 70 studies, of which 49 also included data on resistance rates (see Supplementary Materials for full list of studies). The total number of patients in all studies was 564 355, and the total number of ventilator-days was 1 079 309 with 21 176 episodes of VAP. The median age of patients was 52.0 years (interquartile range, 40.2-57.6 years) and sex was available for 141 354 patients, of whom 93 108 were male (65.9%).

Incidence Density and Prevalence Data
Data on incidence density were available from 14 countries, which included 538 600 patients, 1 076 140 days of mechanical ventilation, and 16 082 episodes of VAP. The majority of these studies were published in the last 10 years (26/31 [84%]). Significant heterogeneity was observed in the studies as a whole (I 2 = 100%) and when the studies were divided both by country and by income group (I 2 approximately 99% in all categories). Therefore, all analysis was done using the random-effects model. The point estimate incidence density of VAP/1000 ventilator-days in all studies was 15.1 (95% confidence interval [CI], 12.1-18.0). Meta-analysis of incidence density per income level is presented in Table 1. There was a higher incidence density rate in LMICs compared with HICs when analyzed by meta-regression (18.5 vs 9.0 per 1000 ventilator-days; P = .035). Meta-analysis of studies by study quality, study type, year of publication, hospital type, and ICU type did not show any significant difference in incidence density or prevalence (see Supplementary Materials for full analysis).
The pooled period prevalence of VAP for all studies was 12.7% (95% CI, 10.0-16.1). This includes 18 400 episodes of VAP and 549 478 patients. Meta-analysis of period prevalence per income level is presented in Table 1 (see Supplementary Materials for full analysis). Again, analysis by study quality, study type, year of publication, hospital type, and ICU type did not show any significant difference. Period prevalence was highest in UMICs and lowest in HICs. Pooled incidence density and pooled prevalence data per country are presented in Table 2, and prevalence by country and region in Figure 2. The highest pooled incidence density was found in Mongolia (43.7/1000 ventilator-days) and the highest pooled prevalence in Hong Kong (48.1%), each based on 1 study [19,20]. When the data were split by income level, A. baumannii was the most common organism in the LMIC group. However, the proportion due to this organism gradually reduced as income levels increased, and S. aureus and P. aeruginosa were the most common in the HIC group ( Figure 3A). When the studies were divided by geographical latitude into tropical and nontropical regions, A. baumannii was more common in the tropical regions whereas S. aureus was a more common cause of infection in nontropical regions ( Figure 3B). Aggregated microbiology data are presented by country for the top 7 organisms in Figure 4.

Cost Burden of VAP
There were very few data available on the costs of VAP in Asia. Only 5 studies in our review included any information on the economic burden of VAP and these used different methodologies and therefore could not be compared. Three studies gave ICU costs of patients with and without VAP. In 2011-2012, in Ludhiana, India, it was estimated that the average cost of an ICU admission for a patient with VAP was $6250.92 compared with $2598.84 for a patient without VAP [21]. In Tehran, Iran, in 2012-2013, VAP cost 3.04 purchasing power parity (PPP) per day, compared with 1.23 PPP per day for those without VAP [22]. In Taipei, Taiwan in 2011-2014, an average ICU admission with VAP cost $8924.50 compared with $6625.50 without [23]. One further study in 2011-2012 in Goa, India, calculated the average additional cost of antibiotics used to treat VAP at $432.7 [24]. An additional study estimated the cost benefit of introducing a VAP educational bundle in Thailand and found the average hospital cost per patient admitted to ICU was $4769 prior to the bundle and $2378 after the bundle [25].

DISCUSSION
VAP remains a major contributor to HAIs in Asia. This systematic review highlights very worrying aspects of the ongoing crisis in HAIs and antibiotic resistance affecting this region. The incidence density of VAP remains high in both LMICs and UMICs, demonstrated by no indication of a reduction over time on meta-analysis. Given the large expansion of ICUs in these settings, the at-risk population can be expected to rise substantially in the immediate future, compounding the problem. Additionally, for LMICs anticipating a resolution of the problem as their economies grow, the data from the UMICs in the region are worrying. There are major omissions in the published data from many of the countries included in this review with only 22 of 40 countries having any published data at all. The changing trend in microbiological etiology from previous published reviews (where S. aureus was the predominant microorganism) and the rate of resistance in these organisms reinforces the picture of a global antibiotic crisis [26,27].

Incidence Density and Prevalence Data
To our knowledge, this is the first review to analyze incidence data by national income level. We have demonstrated a trend in the incidence density of VAP by income level but no trend in the period prevalence. Our pooled period prevalence has demonstrated that UMICs also have a significant burden of VAP and that this may be higher than LMICs. When comparing with other international VAP studies, the International Nosocomial Infection Control Consortium (INICC) report from 2009 gives a pooled density of 13.6 per 1000 ventilator-days, which includes data from Latin America, Asia, Africa, and Europe [28]. Our overall rate is similar to this; however, when taking the LMIC alone, we have shown much higher rates, similar to those documented in a systematic review of developing countries globally (22.9/1000 ventilator-days) [29].
It is currently complicated to compare national rates of VAP with the rate in the United States due to changing definitions and means of classifying VAP [30]. However, the published rates of 0.0-5.8 per 1000 ventilator-days in the United States are comparable to some but not all of the studies from HICs in this review [31]. Previous reviews have quoted incidence density rates of 2-16 per 1000 ventilator-days (from the INICC and United States), whereas we have found much higher rates in some of our studies (0.50-76.83) [32]. A systematic review and meta-analysis of VAP in China determined the incidence density at 24.14 episodes per 1000 ventilator-days [33]. This is higher than the rate we have found in China (19.0/1000 ventilator-days). Whereas we included 13 articles from China, the aforementioned review included 195 additional studies published in non-English-language journals, suggesting higher rates in these publications.

Microbiology and Resistance Rates
This review adds to the evidence of A. baumannii as the primary organism associated with VAP; it is the commonest organism in LMICs and in countries in the tropics and has high rates of antimicrobial resistance in these regions. This is the first systematic review to collate studies from Asia that demonstrates A. baumannii as the main cause of VAP [11,34]. Other studies have demonstrated other gram-negative bacteria, such as P. aeruginosa, as a major cause of VAP [10]. However definitive conclusions are hard to draw since the diagnosis of VAP is complicated and isolated pathogens may be colonizers and not the causative agent.
The resistance rates we have found are similar to those documented by the INICC and others globally and demonstrate a worrying rise in resistance in all pathogens to even the broadest antibiotics (eg, rates of carbapenem resistance in A. baumannii at approximately 75%) [35][36][37][38]. Resistance rates to colistin were not available in any of the studies. Our review confirms what multiple single-site studies and expert reviews have stated, that resistant gram-negative bacteria are the major cause of VAP in Asia [37,39].

Cost Burden of VAP
There is very sparse information on the health economic impact of VAP in Asia. There are several studies describing the disease burden of HAI in Asia or LMIC settings, all of which have found high rates of VAP and a significant increase in mortality and length of stay [16,29,40]. Our study aimed to summarize this sparse literature on direct economic burden of VAP, but due to the lack of published data on this we are unable to draw any definitive conclusions. It would appear that VAP is a significant cause of patient morbidity and mortality, that the rates remain high, and that cost of prevention bundles in this setting are both clinically and financially worthwhile.
Our study is limited by the inclusion of only English-language journals. We excluded several studies from our search due to use of a non-English language and we did not search non-English-language databases. In addition, there was a large degree of heterogeneity in our studies, which we have attempted to compensate for by using the random-effects model for meta-analysis, but the underlying differences in the studies remain. These differences result from the wide geographical area covered as well as different practices within the ICU, different patient groups within the ICUs, and different definitions used to diagnose VAP. Due to the publication bias and heterogeneity, any conclusions and extrapolations to other populations should be undertaken with caution.
In conclusion, VAP is a significant burden in Asia, with LMICs and UMICs both demonstrating higher rates than HICs. There is a worrying amount of antibiotic resistance in the pathogens causing VAP and unless concerted international effort is made, this is likely to worsen.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes
Financial support. This work was supported by the Wellcome Trust through the Asia Program Core Grant.
Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.