Factors Associated With Prolonged Hospital Stay: a Retrospective Study for Imported Malaria Cases

Liang Huang Chengdu Public Health Clinical Center https://orcid.org/0000-0002-9607-7939 Hong Jin Chengdu Public Health Clinical Center Hong Zhang Chengdu Public Health Clinical Center Yang Liu Sichuan Center for Disease Control and Prevention Xinxing Shi Chengdu Public Health Clinical Center Xintong Kang Chengdu Public Health Clinical Center Yilan Zeng Chengdu Public Health Clinical Center Lin Wang (  15201920@qq.com ) Chengdu Public Health Clinical Center

. China has entered the era of post-elimination, all the malaria cases were imported from oversea. Most of them were exported labour to Africa, the steadily increasing number of imported patients has been has been imposing great public risk to China society 5,6 .
Our hospital located in Chengdu City, which is the provincial capital of Sichuan province(Western China, including eastern Qinghai-Tibet Plateau), is the only tertiary hospital for infectious disease in Chengdu City. It is providing services for Chengdu city and most area of the Sichuan province. It hospitalized majority cases of malaria imported into Sichuan provinces. In 2019, our hospital passed the national examinations for NMEP, quali ed for malaria patients managements, diseases reporting, treatments, diagnosis, laboratory things, and staff training. So that the dataset of malaria from our hospital could be a representative subset for imported patients in China.
Treatments for malaria would be successful and effective if appropriate intervenes engaged in time 7,8 .
However, according to our experience, the clinical conditions and length of stays(LOSs) for patients varied in a wide range. In this study, we focused on investigation of the predictive factors for prolonged hospital stay, give the evidence for clinical decisions and patients managements. No previous study had covered this topic for imported malaria cases in China.

Patients
We retrospectively collected malaria cases for Hospital information System(HIS) from 2017 to 2020.
Totally 150 cases we retrieved, we collected clinical descriptions including medical history, main complains, sources of countries, et al. The laboratory tests results including blood routine, liver function, and medical images records.
The written forms of agreements were signed by patients before admissions. Ethical approval for this study was obtained from the Ethics Committee of the Chengdu Public Health Clinical Center. The patient consents for inclusion was waived as the retrospective nature of this study. Criteria All patients con rmed by pathogenic evidence of plasmodium. Patients with suspected malaria, however, excluded by later medical investigation were excluded from this study. The diagnosis was followed with national recommendation for malaria diagnose(WS 259-2015) 9 .
For the criteria of discharge, we routinely followed by rules of our hospital as below: 1. No plasmodium was found in blood by microscopy. 2. Signi cant improvement of clinical signs, manifestations, and laboratory examinations 3. Normal body temperature for more than 3 Days.

Statistical analysis
The statistical analysis was performed with the STATA/SE 14.1 software (StataCorp,4905 Lakeway Drive, College Station, Texas, USA) and R 4.0.2 10 . SEAST module for STATA was installed for Edwards seasonality test 11 . The networkD3 for R was used for diagram plotting 12 .
The continuous data were presented as mean with standard deviations for normal distribution or median with range for other distribution. Multivariable cox proportional hazard regression with stepwise variables selection were used for tting the data. The Edwards Test for seasonality was used for time series data of cases number 11 . Log-rank was used for testing equality of survivor functions. The α was set to 0.05. P < α was considered to be statistically signi cant.

General characteristics of patients
All the patients were Chinese nationalities, they all infected malaria out of China. All of them were exported labor services for Africa or two Asian countries(Myanmar and Pakistan). All of the were male, with the median age of 42 years(range from 19 to 62 ) old. All of them have a healthy background, since they had passed body examinations for being a exported labour before going abroad (Table 1).

Medical history distributions of the patients
In the Hospital Information System(HIS), we recorded the detail of the onset and medical history, and history of current illness. Half of them(50%, 75/150) had been infected by plasmodium for more than once, 68% cases were infected by plasmodium for more than twice.
Of note, 17%(25/150) cases had reported misdiagnosed by other medical facilities, since the malaria lack of speci c clinical signs, many doctors in non-history-endemic area had no experience for diagnosis.
To investigate if there was any delay for diagnosis for any reason, we recorded the days from onset of fever before came to our hospital(DOF). The DOFs were range from 0.2-30 days(DOFs may include the days before custom entry) with the median of 3 days.

Geographical and pathogens distributions
Excepted one case came back from Myanmar diagnosed with P. vivax infection, three cases came back from Pakistan infected by P. vivax, other cases were from Africa. Most of the cases were infected or coinfected by P. falciparum, the proportions were 74%(111/150) and %4(6/150) respectively. The second most common infection were caused by P. vivax. The infection and co-infection proportion were 16% (24/150) and 3%(4/150) respectively.
Geographically, most of imported cases were from Democratic Republic of the Congo, the proportion is 23%(34/150), 74%(26/34) cases were infected by P. falciparum. The second most cases were from Equatorial Guinea, the proportion was 14%(21/150), 81%(17/21) of them were infected by P. falciparum( Table 2,3). A Sankey diagram was used for describe the relationship between source of countries and pathogens (Fig. 1).  We overlapped the variations of cases for the four years in to one chart. We can gured out no signi cant peaks were found. The decreasing of 2020 were due to the due to the polices of quarantine for Covid-19.
To understand if the samples from our hospital has the seasonality. We used Edwards Test for seasonality 11 . The χ 2 = 2.51 p-value = 0.28. This result rejected the signi cant seasonality of the cases admission time distribution from the point of statistical view (Fig. 2).

Laboratory tests of the patients
According to our regulations, all the patients of malaria were undergone emergency medical evaluations regardless the clinical condition and potential prognosis. The assessment included blood routine, blood biochemical tests, plasmodium and RDT 13 , X-ray or CT scan for lung, and in most cases PCT and CRP examinations.

Survival analysis for prolonged hospital stays
The median length of stay(LOS) was 7 days, the range from 3-23 days. We used potential related factors to t into a cox proportional hazard model to investigate the relevant factors and Hazard Ratios(HRs).
The results were listed in Table 1. We de ned the failure event as the discharge following advices(no patient discharge against advices), so that the HR value was the risk for discharge, that was, a risk factor for prolonged LOS if HR < 1.
In this study, pathogen type(dummy variables), bacterial infection, abnormal liver function, thrombocytopenia(< 100*10 9 /L), leukocytopenia(  Table 4. The survival functions were plotted in Fig. 3 and Fig. 4. Of note, according to our analysis, being infected by P. falciparum solely was the risk factor of discharge(HR = 4.93, 95% CI = 1.74-14.00), that was, being infected by P. falciparum solely was the protective factor for prolonged hospital stays. This because of the differences of treatment scheme for pathogens. This was not the major topic of this paper, however, this factor remained in equation to balance the effects of different pathogens.

Treatments and outcomes
The imported cases of malaria had very low mortality which was reported as 0.5%(90/16733) 1 . In our study, all the patients were ful lled the discharge criteria, no patient was dead. All the treatments were complied with national recommendation for malaria treatments(WS/T 485-2016) 14 .

Risk for post-elimination era in China
The imported cases were the major risk for malaria in China post-elimination of malaria, especially imported P. falciparum. Malaria prevention activities should target exported labour groups given the increasing number of workers returning from overseas 6 . The major challenge is the unawareness of malaria for the doctors in primary medical care facilities 15 , the situation could be even worse postelimination. In our study, 17%(25/150) cases reported misdiagnosis by other medical facilities. However, we inputted this factor as an independent variable found it was not statistically signi cant by using cox proportional hazard model. We still not able to reject the negative impact for disease control due to the limitation of the dataset. This remind the continuous effects of training for doctors should not halt in post-elimination era.
Predictive factors for prolonged hospital stay Through cox proportional hazard model analysis, we found the low PLT count and complication of bacterial infection were the predictive factors for prolonged hospital stay. The two factors were protective for patients discharge, that were, the risk factors for prolonged hospital stays.
The prolonged hospital stays were implicated a more serious clinical situation, longer time would be taken to ful ll the discharge criteria. More medical resources would be consumed, and more local transmission risk existed. A previous study showed the PLT count predicted the severity of the disease 16 . Varadaraj had reported 97.6%(82/84) cases of malaria were found to be diagnosed with thrombocytopenia 17 .
Bacterial infection could be common in malaria patients 18 . A systematic review showed that 6.5% infection rate of bacterial was found in Africa Children with P. falciparum 19 . Of note, in our study, all the patients with bacterial infections were bacterial pneumonia and UTI, caused limited consequences after antibiotic treatments. There is still highly possible that blood stream infection occur, it may cause serious problems 20 .

Seasonal patterns of malaria
We had the negative nding of the seasonality by using dataset of all the 150 cases admitted to our hospital during 2017-2020. However, the previous studies for malaria revealed that a seasonal pattern of incidence was found in Burundi 21 . Another study indicated seasonal transmission of malaria in Niger 22 .
Mbouna found the seasonal relationships between climate, population density and malaria indicators in Cameroon 23 . According to our experience and statistical data analysis, we found no seasonality of the cases distributions. We thought the incidence of malaria in Africa could not directly affects the pattern of the exported labors returning, which was normally pre-planned. This nding implicated that hospitals should prepare the special wards regardless the seasonal variations of disease incidences in endemic area. Since in our hospital the wards for malaria patients were isolated from other diseases and equipped with mosquito repelling facilities to eliminate the possibility of local transmission.

Clinical implications for geographical and pathogens distributions
Understanding the geographical distributions of the pathogens plays pivotal roles for clinical diagnosis and treatments. Doctors could deduce the pathogens for patients judged by where they came from, without hours waiting for laboratory testing results for pathogens, giving the possibility of rapid access for treatment. A review con rmed that rapid access for treatment reduced risk of severe disease 24,25 . From our study, most cases came from Democratic Republic of the Congo(DRC), patients from DRC infected by almost all kinds of plasmodium. However, P.falciparum was the most common pathogen(76%, 26/34).

Consent for publication
Not applicable Availability of data and materials The data supporting the ndings of this study are available from the corresponding author LW on request.   Survival functions for bacterial infection with 95% con dential interval.