Economic Burden Attributable to Healthcare-Associated Infections in Tertiary Public Hospitals of Central China: A Multi-Center Case-Control Study

DRGs: Diagnosis Related Groups; HICs: High-income Low- and Middle-income GDP: Gross Domestic Product; aCCI: age-adjusted Charlson ICD: International Classication of Diseases; LOS: Length of stay; NISS: Nosocomial infections surveillance system; IPC: Infection prevention and control; CI: Condence Interval; IQR: Interquartile range; ICU: Intensive care unit; CMI: Commercial medical UEBMI: Urban employee basic medical URBMI: Urban basic medical NRCMI: New rural cooperative medical HSIs-PS: Healthcare service items-payment system; SD-PS: Single disease-payment system; DRGs-PPS: Diagnosis Related Groups-prospective payment system; CLABSI: Central line-associated bloodstream infection; VAP: Ventilator -associated CAUTI: Catheter-associated CRE: Carbapenem-resistant while S is for in CSE, CSPa, and

Furthermore, the economic losses attributable to device-associated infections and hospital-acquired multi-drug resistant bacteria were 2 to 4 times those of the controls.

Conclusions
The economic burden attributable to HAI in central China is heavy, and opportunities for easing this burden exist in several areas, including that strengthening antibiotic stewardship and practicing effective bundle of HAI prevention for patients carrying high risk factors, for example, elders or those with catheterizations in healthcare institutions, and accelerating the medical insurance payment system reform based on Diagnosis Related Groups (DRGs) by policy-making departments.

Background
Healthcare-Associated Infection (HAI) occurs with the founding of hospital, which is characterized by high morbidity and mortality [1]. And it is not only life-threatening and increases burden on individuals and families, but also causes huge resource wastes and economic losses for hospitals and society. The cumulative burden of HAIs was about 501 disability adjusted life years per 100,000 population each year, which was higher than the total burden of all other 32 kinds of diseases included in the Burden of Communicable Diseases Project in Europe [2]. In the United States, 1.7 million people suffer from HAIs every year, which causes an economic loss of $8.3 billion to $11.5 billion [3].The impacts HAI has on patients [4], hospitals [5] and society [6] are well recognized, while most of them focused on high-income countries (HICs). What is worth mentioning, the low-and middle-income countries (LMICs) have limited medical resources but high incidences of HAIs, resulting in relatively larger incidence of patient disability, mortality and additional hospitalization cost [7]. However, the burden attributable to HAIs in LMICs remains poorly de ned compared with that in HICs. Moreover, due to the objective factors vary, such as the demographic and sociological characteristics, medical insurance policies, economic development levels and hospital scales, the existed health economic characteristics of HAIs may not have universal applicability and cannot be generalized to another hospital, country or region as a whole. Incidentally, As one of the most populous and medical resources scarce provinces, Henan Province has about 109 million population, one third of the population of Central China, and 19 million discharged patients in 2018 where there has been no research focusing on exploring the health economic characteristics of HAIs because of the absence of representative data. We therefore conducted a multicenter, retrospective, standardized case-control study, to accurately estimate the current economic burden of HAIs in tertiary public hospitals of Central China, and to provide data support and factual evidence for further research and policy making.

Patients and study design
We adopted a three-stage random sampling method to select patients with HAIs in tertiary public hospitals of Henan Province. In the rst stage, based on the economic level, all 18 cities were ranked by their Gross Domestic Product (GDP) in 2018, and the rst of every three cities was chosen. In the second stage, according to the number of tertiary public hospitals of the included cities and their feasibility of conducting this survey, one or more hospitals were selected in each city by using strati ed sampling. In the last stage, with a systematic sampling strategy, all patients suffered from HAI in the selected hospital between January 1 and December 31 2018 were ranked by their admission numbers, and the rst of every seven patients was selected into the HAI group.
Then we designed the study to have 1:1 matching, with one control who did not suffer from HAI for each case. In order to reduce the confounding bias caused by undermatching or overmatching, controls were selected according to the following matching criteria: (1) the rst discharge diagnoses were same, coded by the International Classi cation of Diseases, 10th Revision (ICD-10, Version: 2016); (2) the age-adjusted Charlson Comorbidity Index (aCCI) were equal; (3) the surgeries undergone were same, coded by the ICD Clinical Modi cation of 9th Revision Operations and Procedures (ICD-9-CM-3); (4) the gender were same; (5) the age gap was 5 years or less, and no more than half a year for children under 5 years old; (6) The inpatient departments were same; and (7) the difference of admission date was a month or less. Patients with length of stay (LOS) ≤ 2 days were excluded, and if there was more than one patient without HAI meeting the above matching criteria, selected the one who had smallest age gap with infected patient into the control group. The HAIs were diagnosed according to the Diagnostic Criteria for Nosocomial Infection, which was published by the National Health Commission of China in 2001 [9]. The study conforms to the ethical principles of the 2013 revised Declaration of Helsinki and received the Ethic Committee approval from all of the surveyed hospitals, with a waiver for patient informed consent.

Data collection
The hospitalization cost, demographic and clinical characteristics of patients were retrieved from the Hospital Information System, and the epidemiological characteristics of HAIs were obtained from the Nosocomial Infections Surveillance System (NISS) of the selected hospitals. The cost of infection prevention and control (IPC) was collected through eld questionnaire surveys, which mainly comprises o ce expenses, labor cost of full-time and part-time staff, NISS maintenance fee, funds of activities such as training, seminar and so on. The discharge diagnoses were retrieved from the home page of electronic medical records, and the aCCI was calculated by weighting each condition to assess the aggregate burden of comorbidity [10]. The detailed calculations of hospitalization cost, IPC cost and economic loss attributable to HAI are shown in the Appendix. The average exchange rate of CNY (¥) to USD ($) was 6.86:1, issued by The People's Bank of China from the period over which the study took place. [11] An investigator-uni ed training was conducted before the survey, and data validation was performed with double entry in the process of data extraction.

Statistical analysis
We used EpiData 3.1 (EpiData Association, Odense, Denmark), Excel 2010 (Microsoft Corporation, Seattle, Washington, USA) for data collection and mining, and SAS 9.4 (SAS Institute, Cary, NC, USA) for data analysis. For continuous variables (i.e. LOS and hospitalization cost) we veri ed the distribution types by using Kolmogorov-Smirnov test and calculating the coe cients of skewness, and then described their central tendency with mean and 95% con dence interval (95% CI) or median and interquartile range (IQR), as appropriate. The Wilcoxon signed-rank test (W test) was adopted to compare the difference of hospitalization costs between matched pairs of patients. Then a subgroup analysis was performed and the Kruskal-Wallis H test or Mann-Whitney U test were used to identify the heterogeneity of economic losses attributable to HAIs among different medical insurance types, payment systems, infection sites and pathogens. In addition, the Spearman rank correlation coe cient was calculated to analyze the correlation between the prevalence of HAI and investment of its prevention, as well as that between the economic loss and patient's age. Considering of the low power of nonparametric test, the signi cance level (a) was set to 0.05, not to 0.01, to reduce the probability of false negative errors.

Characteristics of patients
A total of 2976 patients in 10 hospitals (accounting for 12.99% of all tertiary public hospitals in Henan Province) were enrolled, including 7 hospitals with more than 2000 beds. No signi cant differences were found between the two groups with respect to gender, age, hospitals, aCCI, surgery and admission to ICU (P>0.05; Table 1).

Prevalence of HAI and cost of IPC
The overall incidence rate of HAI in the selected hospitals was 2.42% (range, 0.88% to 4.15%). And the cost of IPC per 1000 beds was $35644.24 (range, $24929.76 to $53146.41), which was signi cantly, but negatively, associated with the incidence rate of HAI (Spearman r=-0.76, P=0.03).

LOS and hospitalization cost
The length of hospital environment exposure prior to the onset of HAI was 8 days (IQR, 3 to 12 days). And the LOS of HAI group was 23 days, which was 10 days (IQR, 8 to 16 days) signi cantly longer than that of control group ( Table 1). The per patient hospitalization cost in HAI group was $2047.07 higher than that in control group. Among hospitalization cost types, the gap of pharmaceutical cost between two groups ranked top with $1044.39 (excess antimicrobial drug cost accounted for 59.77%; Table 2).

Correlation between economic loss and age
The hospitalization cost of HAI patients were signi cantly higher than that of control patients on the corresponding age levels, and there existed a signi cant correlation between the economic loss attributable to HAIs and age (Spearman r=0.26; Table 3).

Economic losses strati ed by medical insurance types and payment systems
The differences of economic losses attributable to HAIs among the subgroups of different medical insurance types had marginal statistical signi cance (P=0.03), and the economic losses in the subgroup of CMI, UEBMI and URBMI were $1834.47, $643.28 and $223.49 higher than the overall median loss, respectively (Table 4). Furthermore, these losses in three different medical insurance payment systems had signi cant difference (P<0.05), too. The economic losses in the subgroup of SD-PS and DRGs-PS were $1135.42 and $1463.63 lower than the overall median loss, respectively (Table 4).

Economic losses strati ed by infection sites
Except the skin and soft tissue, the differences of hospitalization costs between patients with HAI in different infection sites and control group were statistically signi cant. And the most economic losses attributable to HAIs occurred in the hematologic system ($4734.20) and nervous system ($4197.49). In addition, it was worth noting that the economic losses caused by VAP and CAUTI were 4.14 and 2.87 times signi cantly higher than those caused by the other HAIs of the respiratory system and urinary system, respectively (Table 5).

Economic losses strati ed by pathogens
A total of 568 (38.17%) clinical isolates of pathogens were cultured from patients with HAI, and Escherichia coli (18.13%) was the most frequently isolated bacterial, followed by Acinetobacter baumannii (12.68%) and Klebsiella pneumoniae (11.27%). The economic losses attributable to HAIs caused by different pathogens had statistical signi cance, of which Acinetobacter baumannii was on the top list with $9882.75. In addition, the economic losses caused by CRE, CRPa, MRSA and CRAb were 4.06, 3.64, 3.02 and 1.45 times signi cantly higher than those caused by CSE, CSPa, MSSA and CSAb, respectively (Table 6).

Discussion
To our knowledge, this retrospective study is the rst to estimate the current economic burden and analyze the health economic characteristics of HAIs in tertiary public hospitals of Central China. In this work, the estimated economic losses attributable to HAIs was $2047.07, accounting for 28.00% of per capita GDP ($7310.79) and 63.94% of per capita disposable income ($3201.68) in Henan Province, 2018 [8], which is both higher than that of a retrospective survey conducted by Jia HX et al. on 68 general hospitals in China, 2015 [12] and a research did in a referral hospital of Iran, 2017 [13], but lower than the direct economic loss of HAIs estimated by Li H et al. in 5 tertiary public hospitals of Hubei Province, 2016 [14] and that of a similar study made in tertiary hospitals of German, 2015 [15]. On the one hand, it is because the sample size and survey region vary among these studies. On the other hand, by assuming that the economic variables related to hospitalization obey the normal distribution, most of the existing studies used mean as the statistical indicator to describe the central tendency of their distributions [16][17]. Nevertheless, the variables of hospitalization cost and economic loss in our study did not obey the normal distribution, which skewed to the right with a heavy tail, so the statistical indicator of median (lower and upper quartile) was adopted to estimate the economic loss.
In accordance with the results of current researches [17][18][19][20][21][22][23][24][25][26][27][28][29], the subgroup analysis shows that the economic losses caused by VAP and CAUTI were approximately 3 to 4 times higher than those caused by the other HAIs of their corresponding systems, while marginal difference was found when it comes to CLABSI, probably because of the limited sample size and low power of U test. We also found that the economic loss attributable to HAIs came mainly from pharmaceutical cost, of which additional antimicrobial drug cost accounted for about 60%. It could be explained by the fact, that antimicrobial drugs are needed to ght against infections, but along with physician's prescription comes the irrational use of antimicrobial drugs (i.e., using drug under no indication of infection, excessive dosage and overlong duration of treatment) [20], which is an independent risk factor for antimicrobial resistance [21][22]. Meanwhile, the infection of Multiple Drug Resistant Organism (MDRO) not only causes huge economic losses,as our study and other relevant studies [23][24] show, but also increases the irrational and inappropriate use of antimicrobial drugs [24]. Infection and antimicrobial resistance complement each other and come to a vicious circle. Therefore, the result of our study is precisely a reminder of the importance of monitoring drug prescription and controlling drug abuse for the reduction of medical burden and the prevention of MDRO infection.
In addition, this study provides the rst estimate of the HAI burden on patients with different medical insurance types and payment systems, which indicated that, the HAIs occurred in patients who had CMI, UEBMI or URBMI caused huge waste of healthcare resources. It was not surprising, given that the HSIs-PS is still covering most cities of Henan Province. Under this system, the excess hospitalization cost caused by HAI are mostly payed for by the medical insurance institutions and a small remaining part by the patients themselves, while the hospitals do not bear the burden basically. As the result of this study showed,the economic losses attributable to HAIs in HSIs-PS were almost 5 times higher than those in DRGs-PPS, which quanti es payment criteria of different diagnosis related groups classi ed by the complexity of diseases and thus limits the waste of medical resources to some extent. Therefore, some developed countries strongly support the investment of HAI prevention by the medical insurance funds [25], and have established some lists of speci c HAIs that are referred to as "no tolerance" events, thereby reducing the reimbursements to hospitals [26][27].
Our study has several limitations. Considering that the economic burden of HAI includes direct loss of prolonged stay, anti-infection treatment and readmission, as well as the indirect loss which mainly consists of the reduced working hours of family members due to hospital care and the declined labor capacity of patients themselves due to infection and even disability, the total losses attributable to HAIs were underestimated in our research. Moreover, although we con rmed that there was a remarkable negative correlation between the incidence rate of HAI and the cost of its prevention, the cause-and-effect relationship between them cannot be proven by this retrospective case-control study. Further prospective studies are needed to address this issue and validate the importance of maintaining the ongoing nancial investments in HAI prevention and control.
In conclusion, based on a large, retrospective and Henan province population-based surveillance, our study demonstrates that HAIs lead to a great economic loss in tertiary public hospitals of Central China, while reveals the opportunities for easing this burden exist in several areas, including that strengthening the antibiotic stewardship and practicing effective bundle of HAI prevention for patients carrying high risk factors, for example, elders or those with catheterizations in healthcare institutions, and accelerating the medical insurance payment system reform based on DRGs by policy-making departments.

Appendix
(1) Hospitalization cost = the total medical expenses incurred by all hospitalized patients (cost of pharmaceutical + operation + lab test + treatment + examination + blood transfusion + material + bed + nursing care).
And the hospitalization cost per patient = hospitalization cost / the total number of hospitalized patients during this period.
(2) IPC cost = daily expenses of IPC o ce + labor cost of full-time and part-time IPC staff + occupational exposure management expenses + NISS maintenance fee + microbiological monitoring fee of hospital environment + cost of IPC trainings + costs of attending and organizing seminars and academic conferences related to IPC.
(3) The economic loss attributable to HAI = the hospitalization cost of patient with HAI -the hospitalization cost of the corresponding patient in control group.

Consent for publication
Written informed consent for publication was obtained from all participants.

Availability of data and materials
All data analyzed during the study are included in the Tables 1-6.

Competing interests
The authors have declared no con ict of interest.     Abbreviations: CMI, commercial medical insurance; UEBMI, urban employee basic medical insurance; URBMI, urban resident basic medical insurance; NRCMI, new rural cooperative medical insurance; HSIs-PS, healthcare service items-payment system; SD-PS, single disease-payment system; DRGs-PPS, diagnosis related groups-prospective payment system. d With U test, the economic losses in these four subgroups were signi cantly higher than the overall median loss (P<0.05). e With U test, the economic losses in these three subgroups were signi cantly lower than the overall median loss (P<0.05). Table 5. Estimates of economic losses attributable to HAIs strati ed by infection sites