Clinical and financial burden of hospital-acquired infection in China

Abstract

Hospital-acquired infection (HAI) is a critical patient safety challenge globally. In China, with the spread of antimicrobial resistance, the burden of HAI requires imminent attention. Surveillance plays an essential role in the infection control program. However, the traditional surveillance-manual review is time consuming and costly. Based on the manual review, most organizations undertaking HAI surveillance have adopted point prevalence survey or voluntarily report. Such methods cannot comprehensively monitor hospital-wide and long-term HAI burden. In 2010, a realtime nosocomial infection surveillance system (RT-NISS) was implemented in a Beijing hospital (the system is being piloted in ten provinces in China).

Using data from the RT-NISS, this thesis aims to measure the clinical and financial burden of HAI in the Beijing hospital. After conducting a series of systematic reviews and meta-analysis on prevalence, incidence, mortality, and cost of HAIs, I carried out three retrospective studies using HAI and inpatient data from 2013 to 2017 in the site hospital.

As compared with international research, my study showed that the 5-year incidence rate was relatively low (3.6-4.1 per 1000 patient-days). The incidence of respiratory tract infections decreased significantly over time whereas bloodstream and surgical site infections increased significantly. Respiratory tract infection (43.80%) followed by bloodstream infections (BSI) (15.74%), and urinary tract infections (12.69%) accounted for the largest proportion of HAIs. HAIs peaked in the summer among adult and elderly patients.

Due to the increased incidence of BSI indicated by this study, I further investigated on the mortality prediction of BSI with Escherichia coli and Klebsiella pneumoniae. The overall in-hospital mortality rate was 18.67%. Increasing age, presence of a cancer diagnosis, increased severity of illness as measured by the Charlson comorbidity index, use of a ventilator to assist breathing, presence of pathogens which were carbapenem resistant, abnormal temperature frequency (<36℃ or >39℃) during hospitalization, C-reactive protein, white blood cells were significantly associated with BSI mortality. Surgery was a protective factor for BSI mortality. The prediction model had an excellent performance with concordance index of the training and validation set were 0.90 and 0.86, respectively.

Apart from the clinical burden of HAI and BSI, the attributable charge of BSI was also estimated from hospital perspective. I compared the attributable length of stay (LOS) and charges of BSI with three models: conventional non-matching, propensity score matching, and risk set matching. After using matched methods, the mean LOS and charges attributable to BSI both decreased, these values in risk set matching model were higher than those in propensity score matching. In risk set matching model, the excess LOS and charges associated with BSI were 17.35 days and US$16186.62, respectively. It indicated that matching on infection time (risk set matching) might reduce the time-dependent bias.

In summary, the clinical and financial burden of HAIs can be comprehensively estimated with the electronic surveillance data. With expanding the scope of the electronic surveillance system, we can establish the regional and national benchmark for HAI burden, which can provide a comprehensive reference for HAI control and clinical practice in the future.