Proposing a new strategy of combining machine learning with controlling cross-transmission of multi-drug resistant organisms of ICU

Multi-drug resistant organisms (MDROs) has become a global threat to public health. MDROs are normally transmitted from patients to patients via the hands of healthcare workers (HCWs). The key management of MDROs is control dissemination as sooner as possible. Method: We established a predictive rule based simply on experiences, and according to the result of this predictive rule we take a series of precautions of a general intensive care unit (ICU) from January 1, 2018, to December 31, 2019, only in one ward experimentally. In this study, we aim to assess the eciency of the routine care practice which include pre-discrimination by the predictive rule and sequent precautions by doing difference comparisons. After comparing two wards in the hospital expenses and length of ICU stay, there are no statistical differences. Precautions contribute to the association between room number and status of MDROs infection/ colonization(p = 0.033), and infection/colonization rate of MDROs is different statistically between two wards(p = 0.006).


Abstract Background
Multi-drug resistant organisms (MDROs) has become a global threat to public health. MDROs are normally transmitted from patients to patients via the hands of healthcare workers (HCWs). The key management of MDROs is control dissemination as sooner as possible.

Method:
We established a predictive rule based simply on experiences, and according to the result of this predictive rule we take a series of precautions of a general intensive care unit (ICU) from January 1, 2018, to December 31, 2019, only in one ward experimentally. In this study, we aim to assess the e ciency of the routine care practice which include pre-discrimination by the predictive rule and sequent precautions by doing difference comparisons.

Results
After comparing two wards in the hospital expenses and length of ICU stay, there are no statistical differences. Precautions contribute to the association between room number and status of MDROs infection/ colonization(p = 0.033), and infection/colonization rate of MDROs is different statistically between two wards(p = 0.006).
Conclusion the routine care practice had controlled the cross-transmission of MDROs in some extent. Future, studies can engage in updating the predictive model based on big data and referred to experts' experiences and adopt more e cient precautions for strengthening the transmission e ciency.

Background
Multi-drug resistant organisms (MDROs) have become a global threat for public health. In American, more than 700,000 healthcare-associated infections occur annually, and many caused by antibiotic-resistant bacteria [1]. According to statistics from the China Antimicrobial Resistance Surveillance System (CARSS), Methicillin-resistant Staphylococcus aureus (MRSAs) accounted for 30.9% of S. aureus isolates and Carbapenem-resistant Acinetobacter baumannii compose 56.1% of baumannii isolates, respectively, in 2018 [2]. D.J. Morgan and Sarah S. Jackson found that when Healthcare Workers (HCWs) contact patients with MDR bacteria they frequently contaminate their protective gowns and gloves [3] [4]. Multi-drug resistant organisms are normally transmitted from patients to patients via the hands of HCWs [5]. MDROs infection/colonization leads to clinical consequences: intensive care unit (ICU) length of stay, limiting treatment options, increased risk of therapeutic failure, and higher mortality and costs [6][7][8][9]. The key step of decreasing the dissemination rate of MDROs is to cut off the spread chain of people to people.
In China, experts advised that patients should have positive screening once admitted to ICU to nd out whether they have colonized MDROs or not. It aims to take sequence precautions, like isolation, to prevent nosocomial cross-transmission of MDROs as early as possible, since solation measures are recommended to reduce transmission of MDR bacteria in the ICU [5] [10]. But growing bacterial cultures and doing a susceptibility test take a few days in China, usually, it is too late to bene t from isolation.
In this article, we introduce an exploratory experiment carried out in our hospital from January 1, 2018, to now (March 2020). The experiment is about applying an elementary predictive model to pre-discriminate the status of patients before the clinical culture testing results came out and then giving interventions rstly. we also will evaluate the effect of this experiment in this article and raise some opinions in the management of MDROs in ICU.

Hospital setting
This study was undertaken in a general ICU (including ward 2A and 2B) of 32 beds (2A has 12 beds), a university-affiliated hospital, located in Shanghai, China. It is a tertiary hospital, which is divided into two branches: the north one and south one. The experiment practiced in ward 2B of the south one only, because the author (Fei Kaihong) is the head nurse of the south branch's ICU(2B). Patients admitted to ICU are arranged to two wards randomly. In 2B, there have 3 quadruple rooms (bed number from 11 to 14, 19 to 22 and 23 to 26), 2 double rooms (bed number include 15,16,17 and 18), and 6 single rooms.
The exploratory experiment From January 1, 2018, to December 31, 2019, the head nurse of the 2B part applied an elementary predictive rule to pre-discriminate the status of MDROs colonization/infection. There are eight relevant factors in the predictive rule. Meeting one clause, HCWs judged that the patient has a very high risk for colonization/infection-supposing to have colonized/infected MDROs. The detail of this rule shows in table 1. If the status assumed to be colonized/ infected with MDROs, the HCWs will strengthen hand hygiene and wear isolation gowns, before and after contacting patients. If circumstances permitted, everyone predicted positive with MDROs should be isolated into a single room or arrange infected or colonized patients separate from other patients otherwise. This preintervention will continue until acquiring the clinical culture results. We also analyzed the association between room number and MDROs infection status. It is noteworthy that we excluded single rooms in this association analyzation because single rooms are prepared exactly for infected patients to isolated with uninfected patients, there has a strong relevance between the single rooms and infected status-causing bias of our result. The result of the association showed in table 3, we could see that room number is associated with MDROs infection status, p=0.033. It suggested that the exploratory experimental routine practices maybe truly discriminated against colonized/ infected people and the intervention curbed cross-transmission between rooms, so the infection occurred only in some speci c rooms.  [14]. In this study, we did not nd statistical differences between hospital expenses and length of ICU stay.
We analyzed the difference in age and gender between two groups only. But there are lots of factors associated with MDROs status, like invasive operation, having chronic renal disease, APACHE ≥ 15, previous hospital stay ≥ 10 days, emergency surgery, and so on [15] [16].Since these factors can in uence hospital expenses and length of ICU stay, so we got no difference between the two groups. We also think the perfect endpoint of the bene t of precaution is the dissemination rate of MDRO. In our experiment, only patients identi ed by the predictive rule as MDROs infection/colonization having screened the bacteria, so we cannot calculate the cross-transmission rate and assess the predictive rule's discrimination and calibration. A study conducted in Singapore assessed the predictive model of MDROs by area-under-the-curve (AUC), evaluating a not bad result. Because it is a one center study, we could not use that model in our hospital, and the situation is different between countries.
In our study, there does have an association between room number and status. In our knowledge, this is the rst study to analyze the association between room number/ bed number and MDROs infection status. The result suggested that the pre-intervention work and it is possible to control cross-transmission of multi-drug resistant pathogens. Discriminating infection or colonization is just a rst step. The purpose is to control cross-transmission between patients. A study reported percentages of cross-transmission in ICU ranging from 23-53% due to patients' contacts [17]. We know that colonized or infected patients, contaminated hands or grows of HCWs, visitors, and environment all could be reservoirs of MDROs. The WHO suggests HCWs should use contact precautions when providing care to patients who are known or suspected to be infected or colonized with MDROs [18]. The key problem is how to know the patient whether he has been infected or colonized nor not as sooner as possible. A guideline-recommended active surveillance contributes to stopping the spread of MDROs by early detection [19]. N. Yamamoto found that rapid intervention based on a rapid molecular diagnostic assay contributed to reduing nosocomial transmission of carbapenem-resistant Acinetobacter baumannii in ICUs [20]. In this study, we analyzed the length of clinical culture testing, it takes 3.4 days averagely in our hospital. So, we boldly speculated that active surveillance and molecular diagnostic assay still time-consuming. In this article, we want to highlight the necessity of combining machine learning with medicine and the golden time for precaution is patient admission time. Based on big data to gure out predictive models of MDROs is just a kind of combination. A study conducted in China came up with using American Thoracic Society Guidelines to predict the status of MDROs infection or colonization, but it failed [21].
The method of combining machine learning with MDROs status' diagnosis in China is in the initial stage. We innovatively bring forward that using a predictive rule to pre-discriminate the status of MDROs infection or colonization and then guide accurate intervention on admission. Still, there are some limitations: Firstly, because this is a clinical practice, we could not require every patient to screen on admission. Secondly, the sample size is not big enough, the number of dead in ICU is too small. Third, the baseline information is limited, it cannot control the basic situations of patients between patients in 2A and 2B, then, the result of the comparison between two groups was lack of evidence. Most importantly, the predictive model was established empirically. Future we need to work on the perfection of the new model both referring to machine learning and consulting experts in nosocomial infection elds. To evaluate the new model, we should conduct a prospective cohort study or before-and-after study. Discrimination and calibration are essential indexes to evaluate the e ciency of the predictive model. Data on baseline information of patients should collect prospectively as detailed as possible.

Conclusion
In this study, we found that the routine care practice including predicting the status of MDROs and sequent precautions had controlled the cross-transmission of MDROs to some extent. It throws light on that this is a feasible strategy to overcome the challenge of controlling the transmission of nosocomial infection. Future, studies can engage in updating the predictive model based on big data and referred to experts' experiences and adopt more e cient precautions for strengthening the transmission e ciency. Availability of data and materials Authors can con rm that all relevant data are included in the article.

Declarations
Authors' contributions ZQ and FK contributed equally to the writing of this article. ZQ and FK involved in the preparation of the proposal and study design, participated in data collection, data entry and data analysis as manuscript preparation. KM and LX contributed to acquisition of data, conception and design of this study. DM and GM contributed to acquisition, analysis and interpretation of data. ZY contributed to conception and design of this study, have been involved in revising the manuscript critically for important intellectual content. All authors read and approved the nal manuscript.
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Competing interests
The author declare that they have no competing interests.