Incidence of healthcare-associated infections in a tertiary hospital in Beijing, China: results from a real-time surveillance system

Healthcare-associated infections (HAI) are regarded as the most common adverse events in health care service delivery. Evidence indicates HAIs lead to prolonged hospital stay, long-term disability, increased antimicrobial resistance, additional financial burden, and even avoidable deaths [1]. The Study on the Efficiency of Nosocomial Infection Control (SENIC) project demonstrated the importance of surveillance to reduce HAI rates, with data indicating 32% of HAIs could be prevented if all hospitals conducted effective infection surveillance and control programs [2].

The majority of international and national studies estimate the burden of HAIs through two methods: point-prevalence surveys and self-report for targeted infections. Studies from the US, Europe and Singapore showed that HAI point-prevalence ranged from 3.2–11.9% [3‐6]. The US National Healthcare Safety Network (NHSN), German national nosocomial infections surveillance system (KISS), and the International Nosocomial Infection Control Consortium (INICC) reported targeted HAI data, such as: device-associated infections rates [7‐9]. Unlike these two surveillance methods, continuous incidence surveillance can provide real-time information [10]. For the purpose of launching hospital-wide and long-term incidence surveillance with fewer personnel, some different electronic surveillance programs have been developed [10, 11]. The published studies primarily describe the surveillance system components and performance within high-income countries rather than reported HAI incidence [10, 11]. Only a Finland three-year (2011–2013) electronic HAI surveillance study based mainly on antibiotic treatment, indicated HAI incidence, which was 15.8 per 1000 patient-days and 4.9% of all discharged patients [12].

In 2010, a tertiary hospital in Beijing, China developed a real-time nosocomial infection surveillance system (RT-NISS) [13], which collected process data not only from antibiotics administration, but also from multiple hospital information systems, such as microbiological reports, antibiotics administration, diagnosis, and clinical symptoms.

Our primary objective in this study was to describe the incidence trend of all types of HAI and the proportion of specific HAIs over a 5-year study period for the Beijing hospital.

Except for longitudinal trends in HAIs, periodic pattern and seasonal variation of HAIs were observed in some studies [14, 15]. Meanwhile, a regional difference also impacted the seasonal variations, for the surgical wound infections, winter peaks were detected in the USA and summer peaks were reported in Finland [16]. One study in south of China (Guangzhou) reported no obvious seasonal peaks in nosocomial infection [17], however, there exist remarkable variations of the outdoor temperature between south and north in China, no reports of northern China have been published. Thus, our secondary objective was to explore the seasonal variation of HAI incidence.

This study described findings from secondary data analysis of a hospital-wide continuous electronic HAI incidence surveillance system in China. Our results of incidence are similar to a meta-analysis of the HAI point-prevalence in Mainland China, indicating a weighted prevalence of 3.12% (95% CI, 2.94–3.29%) [24]. However, compared to international studies, the incidence rate in our study is relatively low [25, 26]. US researchers estimated the HAI incidence density was 2.6–13 episodes per 1,000 patient-days [25]. One US hospital-wide surveillance reported the incidence rate was 4.59 per 1000 patient days in 2012 [26]. The differences with international rates could be explained by the large denominator (long length of stay) in our study. Due to expansions of insurance coverage to hospital services in 2003 and comprehensive care in 2008 in China, patients preferred to receive in-hospital treatment [27]. Typically, data in 2015 showed that average length of stay in China (9.6 days) was longer than OECD countries (8 days) [28, 29].

Our study documented a much higher proportion of bloodstream infections (15.74%) than the average for China (2.65%) [24], which might be a result of the hospital’s new policy encouraging blood cultures to enhance the detection rate of microbiology test. Even though the rate of respiratory tract infection decreased statistically, the incidence of ventilator-associated pneumonia (7.40 per 1000 patient days in 2017) was obviously higher than US study (1.80 per 1000 patient days in 2012) [26].

For the proportion of device-associated infections, it accounted around only 11–13% of all HAIs in our study, which was lower than published US study (25.38%) [4]. Another US study reported that device-associated infections accounted for 38.7% of pneumonia cases, 62.3% of bloodstream infections, and 77.7% of urinary tract infections [30]. However, ventilator-associated pneumonia and central line-associated bloodstream infections in our study accounted for a much lower proportion of respiratory tract infections (8.17%) and bloodstream infections (11.58%), respectively. The study also showed a low proportion of ICU infections (12.54%), which was apparently lower than the US hospital (33% in 2012) [26].

The current surveillance mainly focused on targeted infections, such as: ICUs, device-associated infections, and surgical site infections. The previous study demonstrated that targeted surveillance missed approximately 50% HAIs [31], our results showed the proportion was even higher. Therefore, our study highlighted the importance of infections control for non-device associated infections and non-ICU inpatient units.

Meanwhile, we observed an overall summer peak in incidence of HAI among adult and elderly. Prior studies showed evidence of seasonal variation with catheter-related bloodstream infections peaking during summer months [32, 33]. Some reports suggested that surgical site infection could be due to a “July effect” explained by staff turnover at teaching institutions [34]. The divergent finding among children may be attributed to the winter peak in respiratory tract infections due to cold weather [35].

A review of HAI electronic surveillance system in developed countries demonstrated that the best performing algorithm included the use of three data sources: positive microbiological results, initiation of antibiotic therapy and diagnosis codes [10]. The RT-NISS synthesized data beyond these three databases to identify suspicious HAI cases. Compared to the golden standard-manual review, validation studies of RT-NISS showed good performance, validating HAIs occurred in July 2011 reported relatively high sensitivity (98.8%) and specificity (93.0%), another validation study reported 71.9% surgical site infections among 3,048 patients could be detected by RT-NISS [18, 36]. Hospital-wide surveillance was ever reported in a US hospital through manual review. However, they invested large human resources, and their annual inpatient admissions was much lower than our hospital [26]. Our electronic surveillance system could provide a reference for healthcare facilities with insufficient human resources.

This study also has some limitations. First, the accuracy of all the data was not validated by manual review of the medical records. However,143,750 suspicious cases (22.67%) were reviewed manually and 0.04% (23361) HAI cases were confirmed by the infection preventionists. Second, The RT-NISS lacks surveillance for surgical site infections after hospital discharge. A systematic review showed that 60% of surgical site infections appeared after discharge²³. The previous comparison between manual review and the RT-NISS showed that 94% were detected by RT-NISS from current admission and readmission and 18% were detected post-discharge¹⁶. Third, owing to the updated screening algorithm of the surveillance system, some data were not available or underestimated at the beginning years. Such as: the data of device-associated infections were not available, and the increasing rate of surgical site infection in 2015 was mainly caused by the system updated, some infections without fever and microbiology test were identified through the antibiotic usage upgrade. Forth, the incidence rates in current study were estimated without any risk adjustment strategies. The importance of risk-adjusted measurement was mentioned in different studies, such as: NHSN risk-adjusted measures for surgical site infection [37], McCabe and Jackson score for survival of patients with gram-negative bacteraemia [38]. We will add the procedure-specific risk adjustment variables and disease severity assessment to the surveillance system in the future. Fifth, lack of data about infection control interventions during the study period, relationships between the incidence rate and interventions should be further explored. As the surveillance system only collected the frequency of antimicrobial usage and their resistance pattern, a lack of detailed drug utilization data in this study, such as: defined daily dose (DDD). More comprehensive information about antimicrobial usage need to be collected in the future surveillance.

This study, even a single center study, shows how a hospital-wide continuous electronic surveillance system based on existing hospital electronic databases could provide a practical means for measuring hospital-wide HAI incidence. Our study showed the incidence slightly decreased over the last five years, and it provided an evidence of the necessity for infection control in bloodstream infections, ventilator associated pneumonia, non-ICU and non-device-associated infections. The summer peak also showed the interventions on specific HAIs should be emphasized.