Health itinerary-related survival of children under-five with severe malaria or bloodstream infection, DR Congo

Background Prompt appropriate treatment reduces mortality of severe febrile illness in sub-Saharan Africa. We studied the health itinerary of children under-five admitted to the hospital with severe febrile illness in a setting endemic for Plasmodium falciparum (Pf) malaria and invasive non-typhoidal Salmonella infections, identified delaying factors and assessed their associations with in-hospital death. Methodology Health itinerary data of this cohort study were collected during 6 months by interviewing caretakers of children (>28 days − <5 years) admitted with suspected bloodstream infection to Kisantu district hospital, DR Congo. The cohort was followed until discharge to assess in-hospital death. Principal findings From 784 enrolled children, 36.1% were admitted >3 days after fever onset. This long health itinerary was more frequent in children with bacterial bloodstream infection (52.9% (63/119)) than in children with severe Pf malaria (31.0% (97/313)). Long health itinerary was associated with in-hospital death (OR = 2.1, p = 0.007) and two thirds of deaths occurred during the first 3 days of admission. Case fatality was higher in bloodstream infection (22.8% (26/114)) compared to severe Pf malaria (2.6%, 8/309). Bloodstream infections were mainly (74.8% (89/119)) caused by non-typhoidal Salmonella. Bloodstream infections occurred in 20/43 children who died in-hospital before possible enrolment and non-typhoidal Salmonella caused 16 out of these 20 bloodstream infections. Delaying factors associated with in-hospital death were consulting traditional, private and/or multiple providers, rural residence, prehospital intravenous therapy, and prehospital overnight stays. Use of antibiotics reserved for hospital use, intravenous therapy and prehospital overnight stays were most frequent in the private sector. Conclusions Long health itineraries delayed appropriate treatment of bloodstream infections in children under-five and were associated with increased in-hospital mortality. Non-typhoidal Salmonella were the main cause of bloodstream infection and had high case fatality. Trial registration NCT04289688


Competing Interests
On behalf of all authors, disclose any competing interests that could be perceived to bias this work.
This statement will be typeset if the manuscript is accepted for publication.
Review the instructions link below and PLOS NTDs' competing interests policy to determine what information must be disclosed at submission.
The authors declare no conflict of interest.

Data Availability
The data presented in this study are available from the corresponding author on request. The data are not publicly available due to an embargo on the data until the data collection, data analysis and manuscript publication of an affiliated study with partially overlapping participants and data collection (DeNTS study,NCT04473768) are finalized.

47
Severe febrile illnesses, including severe malaria and bacterial bloodstream infections, are a major cause 48 of under-five mortality. To save lives, on-time hospital referral is crucial to treat severe febrile illness 49 appropriately as soon as possible. In a district hospital in DR Congo, we observed that one third of 50 children with severe febrile illness had a long health itinerary, which meant that they arrived at the 51 hospital >3 days after fever onset. Long health itinerary was most frequent (50.8%) in children with 52 bloodstream infection. In-hospital death was more frequent in children with bloodstream infection 53 (22.8%) than in children with severe malaria (2.6%). In-hospital death mostly occurred early after 54 hospital arrival (admission day 1-2) and having a long health itinerary was associated with death. Finally, 55 we observed that consulting multiple prehospital healthcare providers and prehospital injections & 56 overnight stays were associated with long health itinerary and in-hospital death. We conclude that 57 children with severe febrile illness often arrived late at the hospital, which might contribute to high case 58 fatality, particularly for children with bloodstream infection. Future research and interventions should 59 focus on improved danger sign recognition and earlier referral of children with bloodstream infections by 60 (in)formal frontline healthcare workers. 61 Introduction 62 In sub-Saharan Africa, severe febrile illness accounts for most of the under-five mortality [1,2] and is a 63 France)) is routinely sampled free-of-charge upon admission in all children suspected of having a 112 bloodstream infection (Supplementary Table S1) [16,17]. Blood culture surveillance has revealed a high 113 burden of non-typhoidal Salmonella (NTS) bloodstream infections with NTS causing three out of four 114 confirmed bloodstream infections in children under-five admitted to Kisantu hospital [17]. 115 Eligibility screening, enrolment, and data collection 116 All children who presented at the paediatric ward were screened for eligibility by a trained study nurse. 117 Reasons for non-eligibility were recorded (Fig 2). Caretakers of eligible children were asked for informed 118 consent before enrolment and data collection. Enrolment and data collection occurred 7/7 days during 119 working hours. Children who presented after working hours were enrolled the next morning. 120 Health itinerary refers to everything that happened with the child between the fever onset and hospital 121 admission (Box 1). The health itinerary of each child was questioned via a semi-structured interview with 122 the caretaker by a trained research nurse or physician as soon as possible after arrival of the child at the 123 hospital. In addition, data on referral of the child were collected from the referral letter. All data were 124 entered directly in an electronic case report form (RedCap, Vanderbilt University, Tennessee, US) on a 125 tablet. 126 Venous blood for malaria microscopy examination and a blood culture were sampled and worked-up as 127 part of routine patient care and blood culture surveillance. Blood culture work-up and pathogen 128 identification occurred on site, according to previously published procedures [16][17][18][19][20]. Rapid malaria 129 antigen tests (SD BIOLINE Malaria Ag P.f./Pan test 05FK60, Abbott, Chicago, US) were inoculated with 130 capillary blood for each enrolled child. Together with the results from malaria microscopy, these tests 131 allow to differentiate current (positive Pf malaria microscopy), recent (negative Pf malaria microscopy, 132 Pf-HRP2 antigen positive and pan-LDH antigen negative) and very recent (negative Pf malaria 133 microscopy, Pf-HRP2 antigen positive and pan-LDH antigen positive) Pf malaria infections [21,22] [14]. The Sankey diagram was created with R-package 'plotly'. In-150 Health itinerary: sequence of all actions of healthcare seeking and care provision (at home or from an external healthcare provider) between fever onset and hospital admission.

Long health itinerary: hospital admission > 3 days after fever onset [61]
Home care: administration of medicines that were still available at home, either remaining from prior illness or bought and stored at home in advance to illness [10] Healthcare seeking: any care sought from an external healthcare provider (briefly referred to as provider) [9] Delayed first healthcare seeking: first healthcare seeking > 1 day after fever onset [7,8,38,62] Formal healthcare seeking: healthcare seeking at registered health posts, health centres or hospital. These providers were grouped as formal healthcare providers [9,10,33,34] . Health posts offer a limited package of curative and preventive healthcare services compared to health centres and are run by nurses or health workers [14].
Informal healthcare seeking: healthcare seeking from private pharmacists or drug vendors (further referred to as pharmacy), traditional healthcare providers (religious practitioners/institutions also considered as traditional), private practitioners or private health centres. These providers were considered as informal healthcare providers [9,10,33,34].
Delayed formal healthcare seeking: formal healthcare seeking > 1 day after fever onset [7] Caretaker: person who sought care with the child at the hospital [9] Overnight stay: Passing the night in a healthcare facility before hospital admission hospital survival per diagnostic subgroup was compared by Kaplan-Meier survival analysis (R-packages 151 'survival' & 'survminer'). Unadjusted associations with long health itinerary and in-hospital death were 152 assessed with logistic regression. Adjusted odds ratios were calculated by multivariable logistic 153 regression per category of delay on a subset of variables with significant unadjusted associations with 154 long health itinerary and in-hospital death respectively to avoid multicollinearity. The manuscript was 155 written according to the STROBE guidelines for cohort reporting (Supplementary Table S2

Study population 164
During the six-month study period, 784 children were enrolled as "first-time" visit. The majority (60.3%, 165 473/784) were <2 years old. Figure 2 provides a detailed sociodemographic profile of the study 166 population. Most (86.4%, 677/784) enrolled children were living in semi-urban health areas (Nkandu,167 Kintanu and Kikonka) surrounding the hospital, which together account for 47% of the population from 168 Kisantu health district (Fig 1). Ten enrolled children were living outside of Kisantu health district. Complete results of malaria microscopy and rapid diagnostic test were available for 99.4% (779/784) of 176 children. For five children microscopical Plasmodium species identification was missing but all were 177 considered as Pf malaria based on Pf-HRP2 antigen positivity on rapid diagnostic test. Current malaria 178 infection was diagnosed in 60.6% (475/784) and almost exclusively caused by Pf (Fig 2). Other 179 Plasmodium species were P. ovale (n = 4), P. malariae (n = 3) and P. falciparum -malariae mixed 180 infection (n = 2). Based on the result of the malaria rapid diagnostic test, almost half (45.7%, 139/304) of 181 children with a negative malaria microscopy test had recent (n = 120) or very recent (n = 19) Pf malaria 182 infection. A NTS -Pf co-infection was diagnosed in 8.0% (61/763) of children (Fig 2) and accounted for 183 6.0% of children with current Pf malaria and 51.3% of children with bloodstream infection. 184 The median duration of health itinerary was 3 days (P25 − P75: 2 -4 days) and 36.1% (283/784) of 199 children had a long health itinerary duration (Fig 3). A long health itinerary was more prevalent in 200 children with blood culture confirmed bloodstream infections (52.9%, 63/119) than in children without 201 confirmed bloodstream infection (33.7%, 218/647; p<0.001; Fig 3). Health itinerary duration did not vary 202 seasonally (median and P25 -P75 equal in wet versus dry season, p = 0.77). 203 Children visited a median of two different providers, accounting for a median number of three 212 prehospital visits per child (P25 -P75: 2 -3 visits). However, the median number of visits to a formal 213 provider was only one (Fig 4). A minority of children never sought help from a formal provider (3.6%, 214 28/784) and almost all (94.5%, 741/784) children were referred to the hospital by a formal provider. 215 Half of the children received home care, from whom only five did not consult a healthcare provider 216 before hospital admission. Two children did not receive home care, nor any other prehospital care. 217 Health centres were consulted by 95.4% of children and accounted for half of the healthcare visits, while 218 health posts were rarely consulted. Pharmacies and traditional practitioners were the most consulted 219 informal providers and were each visited by half of the enrolled children (Fig 4). Private practitioners and 220 private health centres were together consulted by 125 (15.9%) children. 221 Home care and consultation of pharmacies largely occurred at fever onset. Health posts, private health 222 centres and private practitioners were consulted earlier than health centres and traditional practitioners 223  The majority (56.5%, 443/784) of caretakers reported to have consulted a first healthcare provider on 232 the day of fever onset. The minority (19.1%, 150/784) with delayed first care seeking had a significantly 233 longer health itinerary (Table 1). Seeking care from a formal provider was delayed in 60.5% (474/784) of 234 children and these children had a significantly longer health itinerary (Table 1). 235 Children with multiple prehospital visits to healthcare providers and those who visited multiple providers 236 had longer health itineraries (Table 1). Visiting an informal healthcare provider was also associated with 237 a long health itinerary, particularly in the case of visits to private health centres and less in the case of 238 traditional practitioners (Table 1). When adjusted for each other, delayed first care seeking, delayed or 239 no formal care seeking, and the number of visits were still significantly associated with a long health 240 itinerary (Supplementary Table S3). 241 Transport as 2nd factor of delay: mostly rapid and direct transport from referring health centre 242 Referral and admission occurred on the same day for all referred children ( Almost all (95.2%, 746/784) children arrived at the hospital within 2 hours of transport, half of them 247 (372/746) arrived within 1 hour (Fig 1). Distributions of transport time and mode were similar in dry and 248 rainy months. Residence in a rural village and transport time of more than one hour to the hospital were 249 significant risk factors for a long health itinerary (Table 1). 250 Prehospital patient management as 3rd factor of delay: high proportions of antibiotics, intravenous 251 medication, and overnight stays 252 Overall, almost all children received antipyretics before hospital admission and approximately one third 253 received antibiotics or antimalarials (Fig 5). Out of 290 children who received prehospital antibiotics, 254 77.6% (225/290) received Access antibiotics, 17.9% (52/290) received Watch antibiotics and 11.0% 255 (32/290) received antibiotics that were not on the national Essential Medicines List. Watch antibiotics 256 were most frequently used in private health centres (Fig 5). We refer to Supplementary Children who received medication as home care (n = 384) were mainly given antipyretics (96%, 370/384) 259 and rarely antibiotics (4.9%, 19/384) or antimalarials (2.6%, 10/384). Pharmacies mostly sold antipyretics 260 (85.0%, 363/427), but also vended antibiotics and antimalarials to 15.7% (67/427)  Prehospital antibiotics, antimalarials, blood transfusion and fluid therapy were associated with a long 280 itinerary, as did intravenous treatment in general and intravenous antibiotics (Table 1). Children with 281 prehospital overnight stays and children who received prehospital diagnostic blood tests had also more 282 frequently a long health itinerary (Table 1). When adjusted for each other, prehospital antibiotics, 283 intravenous treatment, overnight stays, and diagnostics blood tests remained significantly associated 284 with a long health itinerary (Supplementary Table S3). 285 Health-itinerary related in-hospital survival 286 A long health itinerary was associated with in-hospital death 287 Twenty-one children were lost-to-follow-up due to evasion from the hospital (n = 16) or referral to a 288 specialized service (nephrology: n = 4; surgery: n = 1). Overall case fatality was 7.5% (57/763) and a large 289 difference in case fatality was observed between children with bloodstream infection (22.8%) vs. severe 290 Pf malaria (2.6%) (Fig 6). Two thirds of children dying in the hospital died during the first three days of 291 their hospital admission (Fig 6). Furthermore, eight children died on the way to the hospital and 43 292 children died at the hospital before study enrolment (lag time between admission and enrolment was 293 maximum 15 hours on weekdays, maximum 20 hours on weekend days). From the latter, bloodstream 294 infections were confirmed in 20 children, including 16 NTS bloodstream infections. 295 Overall, a long health itinerary was a significant risk factor for in-hospital death (OR = 2.1, p = 0.007; 296 Figure 6). All six children with NTS bloodstream infection who died had a long health itinerary, while a 297 long health itinerary was less frequent in children who died with severe Pf malaria (50%, 4/8) or with NTS 298 -Pf co-infection (61.5%, 8/13). There was a gradual increase in case fatality according to the length of 299 the health itinerary, up to a case fatality rate of 13.1% for children with a health itinerary of more than 5 300 days. Unexpectedly, this increase in case fatality was not concentrated in the first three days of hospital 301 admission ( Supplementary Fig S1). 302

Informal healthcare seeking, prehospital intravenous treatment & overnight stays were associated with in-303
hospital death 304 In addition to their association with long health itinerary, younger age and residence in a rural village 305 were associated to in-hospital death (Table 2), even when adjusted for the association between long 306 health itinerary and in-hospital death (Table 2) and for each other (Supplementary Table S5). 307 308 More visits and more different providers were risk factors for in-hospital death (Table 2). In-hospital 309 death was associated with having visited a traditional practitioner (Table 2), private health centre (Table  310 2) or health post (unadjusted OR = 5.56, 95% CI: 2.53 -12.2, p <0.001), irrespective of the duration of 311 health itinerary. However, when regressed together, only the association between visiting a health post 312 and in-hospital death remained significant (Supplementary Table S5). 313 314 Prehospital blood transfusion, fluid therapy, intravenous treatment, and intravenous antibiotics were 315 also associated with in-hospital death. Although not associated with a long health itinerary, having 316 received systemic traditional care was significantly associated with in-hospital death (unadjusted OR = 317 1.93, 95% CI: 1.07 -3.49, p = 0.03). Finally, prehospital overnight stay was strongly associated with in-318 hospital death, even when regressed with all other prehospital patient management factors associated 319 with long health itinerary and in-hospital death (Supplementary Table S5). 320    was associated with delayed care seeking, consultation of multiple and informal healthcare providers, 354 living in a rural village and transport times of >1 hour to reach the hospital. A long health itinerary 355 coincided with high proportions of prehospital antibiotics, intravenous therapy (including antibiotics, 356 fluids, and blood transfusions) and overnight stays. One in five children with a bloodstream infection 357 died in the hospital, compared to 1 in 50 children with severe Pf malaria. Having a long health itinerary 358 was associated with twice the odds at in-hospital death. Consultation of multiple healthcare providers, 359 consultation of private health centres, traditional practitioners or health posts, intravenous therapy and 360 overnight stays before hospital admission were associated with higher in-hospital case fatality. 361

Limitations and strengths 362
The single-centre, hospital-based design of this study has its limitations. Even within DR Congo, 363 healthcare utilization differs between districts [26].The flat fee system in Kisantu district facilitates 364 healthcare utilization as demonstrated by the fact that half of the children had already been admitted in 365 the previous 12 months and limits the share of the private prehospital sector [14]. Furthermore, COVID-366 19 related changes in healthcare seeking might have occurred. Nevertheless, the study revealed the 367 impact of public health issues that are widespread in sub-Saharan Africa, e.g. the mixed market of 368 healthcare providers, poor healthcare access in rural areas, inappropriate antibiotic use and unsafe 369 injection practices [9,10,27]. By enrolling hospital-admitted children, we only studied the tip of the 370 iceberg and missed children with severe febrile illness who never sought care, who were never referred 371 to the hospital, who remained in the informal sector or who died before admission. Our results are 372 therefore likely to be a too optimistic sketch of the health itinerary of children with bloodstream 373 infections. Nevertheless, due to the embedment of blood culture surveillance in the routine patient care, 374 we could demonstrate that 20 children who died before enrolment had a bloodstream infection. When 375 taking these children into account, the case fatality of bloodstream infections increased from 21.8% 376 (26/119) to 33.1% (46/139), with the largest part of increased case fatality caused by NTS. This 377 observation showed the importance of early enrolment for accurate case fatality estimates but also 378 points to an underestimation of case fatality, in particular for NTS bloodstream infections. 379 The hospital-based recruitment also had its strengths: it resulted in a high enrolment rate and allowed us 380 to focus on the severely ill population for which timely appropriate treatment is most crucial, to 381 differentiate severe Pf malaria and bloodstream infections, and to link health itinerary to in-hospital 382 survival. Data collection by interviewing caretakers is prone to recall and social desirability bias [28], 383 although this was mitigated by organising the interview as soon as possible after arrival and by an 384 independent and trained study team that was not involved in patient management. Finally, we did not 385 perform any socio-economic or qualitative assessment, did not assess the promptness and 386 appropriateness of treatment upon hospital arrival, and did not study disease recognition, although the 387 latter might be less of an issue for severely ill children [29]. 388 Long health itinerary and high case fatality in children with bloodstream infection 389 As in this study, a long duration of health itinerary in a large proportion of children was observed in other 390 health facility-based febrile illness studies in sub-Saharan Africa [11,[30][31][32][33]. Furthermore, previous 391 verbal and social autopsy studies also revealed long health itineraries in deceased children, e.g. >40% of 392 non-survivors being ill >3 days before they arrived at a health facility [11,30,34]. 393 Two thirds of deaths occurred during the first 2 days of admission and the total number of children who 394 arrived dead at the hospital or died before enrolment (n = 51) was almost equal to the total number of 395 deaths in enrolled children (n = 57). Although some of these children might have been saved with good 396 resuscitation practices, earlier presentation at the hospital is essential to allow prompt appropriate 397 treatment and to reduce case fatality [8]. 398 Case fatality of (NTS) bloodstream infections was ten times higher than for severe Pf malaria and (NTS) 399 bloodstream infections were very frequent in children who died before enrolment. Bloodstream 400 infections are a common and frequently fatal cause of severe febrile illness in sub-Saharan African 401 children. Children with NTS bloodstream infection had more often a long health itinerary than children 402 with NTS -Pf coinfection or severe Pf malaria, which suggests that non-malaria severe febrile illness is 403 poorly recognized in prehospital care. Most children indeed had prehospital blood tests which will have 404 facilitated malaria diagnosis but cannot detect bloodstream infections. Strikingly, none of the private 405 practitioners performed blood tests, which contrasts with WHO recommendations for parasitological 406 testing for all febrile patients in a malaria endemic setting [7]. Because diagnosis of bloodstream 407 infections requires microbiological laboratory facilities that are absent at primary care level, efforts to 408 accelerate appropriate (antibiotic) treatment of bloodstream infections should focus on the recognition 409 of danger signs by frontline healthcare workers and early referral [4,35,36]. 410 Delayed formal care seeking and the mixed market of formal and informal healthcare providers 411 While first care seeking was prompt in most children, consultation of a formal provider was delayed in 412 more than half of the children and was associated with in-hospital death. First care seeking often 413 entailed buying antipyretics at the pharmacy. This practice has been previously recognized and linked to 414 better access, lower costs, and good customer care, but does not appear to prolong the health itinerary 415 [10,32,[37][38][39]. Multiple visits and consultation of multiple providers prolonged the health itinerary. As 416 previously described, caretakers hop from one provider to another if symptoms do not rapidly improve 417 [37,40,41]. This causes interrupted treatment and increases the risk at death [11,30,37]. 418 Consultation of traditional practitioners and private health centres prolonged the health itinerary and 419 was associated with increased mortality. This contrasts with the community's perception of better 420 quality of care in private centres [9], although children for whom a private centre or traditional 421 practitioner have been consulted might have been more severely ill. In sub-Saharan Africa, there is a 422 mixed market of private, public, and traditional healthcare providers and children often receive both 423 biomedical and traditional services during the same febrile illness episode [9,[40][41][42][43][44], as was observed in 424 the confounding associations of traditional practitioners, private health centres and health posts with in-425 hospital death (Supplementary Table S5). 426 The high referral rates demonstrated that the flat fee system in Kisantu hospital encouraged referral and 427 therefore consultation of at least one formal provider [14]. Nevertheless, as was observed in free 428 healthcare initiatives [31,45,46], also non-financial barriers (quality of care, access, acceptability, etc.) 429 should be addressed to improve prehospital care and accelerate the health itinerary [9,10,30,36,43,47]. 430 Furthermore, healthcare workers at Kisantu hospital informally declared having noted abuses of the 431 referral system, such as paid referral letters. Fake names of referring centres and a lack of data of 432 referred patients in the health files of the referring centre were also noticed. 433 The complexity of transport delays 434 In agreement with other paediatric febrile illness studies, health itineraries were longer in children from 435 rural villages or in children who travelled >1 hour to reach the hospital [9,[48][49][50][51]. In rural villages, 436 families are poorer, children more vulnerable due to frequent exposure to Pf malaria, other infections 437 and malnutrition and healthcare services more limited. Children from the most rural and remote health 438 areas did not figure among the enrolled patients, which might explain why rural residence predisposed 439 to in-hospital death, but longer transport times did not. There was no impact of the rainy season on 440 transport times and health itinerary duration. This might be explained by a decision not to go to the 441 hospital during heavy rains due to high transport costs, safety issues or the absence of circulating moto 442 taxis [51,52]. 443 Delayed appropriate treatment due to prehospital antibiotics and injectable medicines 444 A third of the children received prehospital antibiotics and this was associated with a long health 445 itinerary. The proportion of antibiotic use is comparable to national Demographic Health Survey (DHS) 446 data in ill children under-five and is an average score for a low-or middle-income country [27]. As in a 447 healthcare exit survey in Kisantu health district, the proportion of children who received antibiotics was 448 highest for visits to private healthcare providers [26]. The proportion of patients who received antibiotics 449 at health centres in the current study (21.7%) is not representative for the antibiotic use in health 450 centres in general (51.1% in the previous survey [26]), because upon their referral to the hospital more 451 than half of children did not receive treatment at the health centre. The proportion of antibiotic use in 452 pharmacies (15.7%) was much lower than in the previous survey (48.8%), but the latter was based on 453 paediatric and adult records [26]. 454 More than 1 in 10 children received intravenous antibiotics before hospital admission and this practice 455 was associated with a long health itinerary and in-hospital death. According to the local protocol for 456 paediatric patient management at primary healthcare level ( where diagnostic and therapeutic facilities are insufficient to manage bloodstream infections [5,35,36]. 461 Intravenous therapy furthermore carries an important risk at fatal healthcare-associated infections 462 [53,54,56,57], as observed in this study. The fact that more than half of the children received 463 intramuscular antipyretics (with the observation of gluteal abscesses along the study course 464 (unpublished data)) further illustrate the widespread problem of unnecessary and unsafe injection 465 practices in sub-Saharan Africa [58][59][60]. 466 The high share of the private healthcare sector in DR Congo carries a public health risk [10,26]. In 467 addition to the higher total antibiotic consumption in private health centres, more Watch group and 468 intravenous antibiotics were consumed in private centres in the current study and in a previous survey in 469 Kisantu [33]. Caretakers often demand "strong" medication, i.e. injectables and antibiotics [9,58,59], 470 which is given more frequently in private health centres. Similar observations of high-risk practices were 471 made in health posts, which might explain the increased risk at in-hospital death after having visited a 472 health post. Regulation, inspection and quality monitoring of the private sector and health posts in 473 combination with community health education to reduce prescription pressure are required to turn the 474 tide and ensure rational antibiotic use and safe drug administration. 475

476
One third of children under-five with severe febrile illness had fever for more than three days before 477 their admission to Kisantu district hospital. Having a health itinerary of more than three days doubled the 478 odds at in-hospital death. The fact that two-thirds of in-hospital deaths occurred before day three of 479 hospital admission increased the biological plausibility that this was a causal association. Delaying factors 480 associated with in-hospital death were healthcare seeking from traditional, private and/or multiple 481 providers, rural residence, prehospital intravenous therapy ( Data Availability:

500
The data presented in this study are available from the corresponding author on request. The data are 501 not publicly available due to an embargo on the data until the data collection, data analysis and 502 manuscript publication of an affiliated study with partially overlapping participants and data collection 503 (DeNTS study, NCT04473768) are finalized. 504 Author contributions:

Results
Participants #13a Report numbers of individuals at each stage of study-eg numbers potentially eligible, examined for eligibility, confirmed eligible, included in the study, completing follow-up, and analysed. Give information separately for for exposed and unexposed groups if applicable. Descriptive data #14a Give characteristics of study participants (eg demographic, clinical, social) and information on exposures and potential confounders. Give information separately for exposed and unexposed groups if applicable. Outcome data #15 Report numbers of outcome events or summary measures over time. Give information separately for exposed and unexposed groups if applicable.

Figures and result section
Main results #16a Give unadjusted estimates and, if applicable, confounder-adjusted estimates and their precision (eg, 95% confidence interval). Make clear which confounders were adjusted for and why they were included  Supplementary Table S5. Multivariable logistic regression to assess the association between in-hospital 747 death and sociodemographic characteristics or factors from the three-delay model (healthcare seeking, 748 transport and prehospital patient management). To prevent multicollinearity, only variables that were 749 significantly associated with in-hospital death when adjusted for long health itinerary were selected (see 750