Other*: Stockbreeders, Tailors, Shopkeeper Teacher.

A total of 63 patients were included. The mean age was 49.7 ± 18.99 years. Males predominated, with a sex ratio of 2.2.

The hospital frequency of upper digestive hemorrhage was 8.99% (63/701 hospitalized patients).

Farmers and housewives accounted for 36.5% (n = 43) and 28.6% (n = 18) respectively. Other professions (breeders, tailors, shopkeepers and teachers) accounted for 34.9% (n = 22).

Medical history was much more frequent, with 84.2% (n = 42) including 50.8% (n = 32) for epigastric pain, 17.5% (n = 11) for jaundice, 9.5% (n = 6) for liver disease, 4.8% (n = 3) for hypertension and 1.6% (n = 1) for digestive bleeding. Surgical history accounted for 11.1% (n = 7).

The mode of onset of bleeding was hematemesis in 34 cases (54%), melena in 16 cases (20.6%) and anemia in 6 cases (9.5%).

The time between onset of the first signs and admission to hospital was less than 24 hours in 4.8% (n = 3), between 24 and 48 hours in 36.5% (n = 23), between 48 and 72 hours in 31.7% (n = 20) and more than 72 hours in 27% (n = 17).

Clinical symptoms were marked by melena in 96.8% (n = 61), tachycardia in 74.6% (n = 47), conjunctival pallor in 60.3% (n = 38), epigastric tenderness in 55.6% (n = 35), ascites in 33, 3% (n = 21), hepatomegaly in 28.7% (n = 18), altered general condition in 23.8% (n = 15), collateral venous circulation in 20.5% (n = 13), arterial hypotension in 17.4% (n = 11) and jaundice in 2.7% (n = 8).

Hemoglobin levels at admission were below 10 and 7 g/dl in 25.4% and 60.3% of patients, respectively (n = 16 and 38). Only 9 patients (14.3%) had hemoglobin levels above 10 g/dl. Creatinemia was normal in 79.4% (n = 50). Viral markers assayed in 59 patients revealed positive hepatitis B virus antigen (Hbs Ag) in 25 (42.4%) and positive hepatitis C virus antibody (HCV Ac) in 2 (3.4%).

Oesogastroduodenal fibroscopy was performed in 40 cases (63.5%), with a delay of more than 72 hours in 60% (n = 24) from the first bleeding episode. The types of endoscopic lesions were esophageal varices (37.5%), gastric ulcers (25%), gastric tumors (20%), esophagitis (2.5%), erosive gastropathies and erythematous gastropathies in 7.5% each ( Table 2 ).

All patients were hospitalized, with a length of stay of over 72 hours in 84.1% (n = 53).

Blood transfusions were administered in 43 patients (68.3%), antibiotics in 52.4% (n = 33), proton pump inhibitors in 85.7% (n = 54), antihypertensive (beta-blockers) in 22.2% (n = 14), macromolecules in 12.7% (n = 8) and hemostasis endoscopy in 3.17% (n = 2).

Continued

The prognosis of patients undergoing treatment was marked by hemorrhagic arrest in 27 patients (79.4%), recurrence of hemorrhage in 20.6% (n = 7) and death in 12 cases (19%). The main cause of death was hemorrhagic shock, accounting for 58.3% (n = 7).

Annotation: HA*: Hemorrhagic arrest; HR*: Hemorrhagic recurrence.

Statistical analysis of prognostic factors was limited to univariate analysis. This analysis shows a statistically significant association between the evolution and certain prognostic factors such as diabetes comorbidity (P < 0.03), hemorrhagic shock (P < 0.007), encephalitis (P < 0.02) and rural residence (P < 0.001).

The hospital frequency of UDH recorded in our series was higher than those reported by some authors [9] - [11] . This difference could be explained, on the one hand, by the under-frequentation of the regional hospital where this study took place and, on the other hand, by the smaller size of our study sample linked to the underestimation of the frequency of UDH in our regions. Indeed, UDH are considered pathologies linked to bad spells according to certain traditions, which leads communities to seek traditional solutions rather than resorting to health services.

The average age of our patients was close to the average age reported in the Western series [12] - [14] , but higher than those reported in certain studies carried out in Mali [9] [10] and elsewhere in Africa [7] [15] - [17] . As in several studies in the literature [13] [17] - [20] , males predominated in our study. This predominance of males could be explained by the fact that men are more exposed to stress linked to unemployment, poverty and family responsibilities. To cope, they often resort to different types of products such as narcotics, traditional stimulants and other gastro-aggressive products from multiple sources, which are established risk factors for the occurrence of upper GI hemorrhage [18] [21] . The socio-professional strata most affected were farmers and housewives, who had difficulty in accessing care due to poverty, ignorance of the disease, low levels of education and self-medication, which is a factor in aggravation and delayed diagnosis. This same finding has been reported in a number of studies [22] [23] .

In our series, as in those reported by some authors, hematemesis and melena were the main reasons for consultation [15] [22] [23] . This indicates irregular use of health facilities by the population, or late recourse to health services due to poverty or low level of knowledge of the pathology.

Concerning the results of complementary examinations, more than half of the patients had severe anemia on admission, with Hb levels below 7 g/dl.

This high proportion of anemia is partly explained by patients’ late recourse to healthcare. This rate of severe anemia is higher than those reported in some African studies [22] - [24] . Endoscopic examination revealed that the main etiological lesions were esophageal varices and peptic ulcer disease, which is in line with some of the data in the current literature [10] [11] . Over the past decade, the incidence of ulcerative digestive bleeding has fallen by more than 20% [25] [26] , while therapeutic methods have evolved remarkably. However, the causes reported in our study differed from those reported by some authors, where peptic or gastric ulcers were the predominant etiologies [22] [24] [25] [27] - [29] . The time taken to perform a digestive endoscopy was over 3 days. This is close to that reported by Sombié et al. in Burkina Faso (24). Turnaround time is an important factor in the diagnostic profitability of endoscopy, since the earlier it is performed, the more effective it is. The delay observed in performing endoscopy in our study was attributable in the majority of cases to the patients’ behavior or their parents’. Indeed, in the general context of our country, patients first begin the treatment with traditional therapists during the first episodes of digestive bleeding and it is only in the event of failure that they seek treatment from modern medicine and generally beyond the first 72 hours. This unfortunate observation whose explanation would be linked to the social interpretation of digestive hemorrhage considered as an act of bad fate. This fact requires the implementation of holistic awareness and communication strategies for social and behavior change at all levels (community, health structures) so that populations understand the importance of early recourse to health care and the dangers linked to self-medication.

Therapeutically, almost all patients (54/63) had access to proton pump inhibitors. This approach to the pharmacological treatment of non-variceal hemorrhage seems rational to us since it was based on the number of cases recorded and is in line with the literature. Hemostasis endoscopy for variceal hemorrhage was rarely performed, despite its proven efficacy. Only two out of fifteen patients suffering from this condition benefited from this method of management. This discrepancy can be explained by the precarious availability of hemostasis devices (esophageal variceal ligation kits, hemostatic glues, clips, plasma-silver, vasoactive substances etc.), frequent breakdowns of the endoscope and the lack of financial resources of patients who were for the most part not enrolled in compulsory health insurance.

As for antibiotic prophylaxis, more than half of patients had access to it, and it was in line with indications which was not the case in the study by Bigoumba at all, where antibiotic prophylaxis in cirrhotic came up against the presumed risk of germ selection expressed by some practitioners [15] . In compliance with contraindications, beta-blocker treatment was prescribed for cirrhotic patients. Forty-three out of sixty-three patients (68.3%) were transfused. This high transfusion rate can be explained by the high rate of severe anemia in our series, where 60.3% of patients had an Hb level of less than 7 g/dl on admission. Because of the non-permanent availability of red blood cells in the hospital blood bank, whole blood was very often prescribed. Systematic calculation of the Rockall score as part of the clinical routine in emergency medicine in our context would help prevent the deleterious effects of delayed or absent blood transfusion and correlate with the need for transfusion.

Patient outcome was marked by cessation of bleeding in 79.7%, hemorrhagic recurrence in 20.6% and death in 19%. The rates of hemorrhagic recurrence and death observed in our series are comparable to those reported by Diarra M et al. and Koumaré M et al. [10] [11] , but significantly higher than those of several other authors in studies conducted in Africa and elsewhere [7] [15] [17] [22] - [24] [28] [30] . Our high proportion of hemorrhagic recurrence and death could be explained by several factors, including the high number of cases of hemorrhage due to rupture of esophageal varices, renowned for its high lethality in countries where the technical platform is not up to scratch, the high proportion of hyponatremia, comorbidity, late recourse of patients to health services and the high number of cases of cirrhosis in our series. The causes of death were hemorrhagic shock and encephalopathy, in line with the literature [11] [31] .

Statistical analysis of prognostic factors was limited to univariate analysis. This analysis shows a statistically significant association between the evolution and certain prognostic factors such as diabetes comorbidity (P < 0.03), hemorrhagic shock (P < 0.007), encephalitis (P < 0.02), sex male (P < 0.02) and rural residence (P < 0.001). These results are close to those reported by Chaabane et al., Sombié et al. and Razafimahefa et al. where hemorrhagic shock, encephalitis, comorbidities and male sex were associated with the evolution of the disease, particularly the occurrence of deaths and Hemorrhagic recurrence [18] [24] [27] .