In this study stroke accounted 16.5 % of total medical admissions and 23.6 % of the total medical cases of in hospital mortality. This admission rate was higher than findings from Gambia in which stroke patients constitute 5%  and in southwestern Nigeria made up of 4.5% medical admission. However, this finding was in agreement with previous study conducted by Deresse B and Shaweno D in Hawasa in which stroke accounted for 13.7% of all medical admissions . The elevated number of stroke admissions in Ethiopia might be due to lack of awareness, poor risk factor control and being hospital based study with referral bias.
From the total stroke patients admitted, 91(78.4%) patients were discharged alive and around one fifth (21.6%) of them experienced in hospital stroke mortality. From those discharged alive, more than half (57.8%) of the patients were discharged with improvement which was lower as compared to study by Masood et al in Pakistan (91%) , Jowi et al in kenya (93.8%)  and Tirschwell etal in Vietnam (65.8%) , but higher than study done by Gebremariam etal in Ethiopia (47.9%) . On the other hand, our finding was comparable with study done in Ethiopia by Greffie etal in which 59.18% of the patients were discharged with improvement . The outcome of the patients during discharge might vary with the severity of stroke, set up of the hospital, complications, comorbidities and experts available in caring of the patients.
The median length of hospital stay was 9.21days which was shorter than previous studies by Walker etal 19 days , Jowi et al 12.5 days , Greffie etal 13 days , Gebremariam etal 11 days  and De Carvalho etal 15.4 days . There are couple of reasons for the shorter length of hospital stay of stroke patients in our study setup. First, some patients were rapidly improved and discharged because they get better quality of care during the early phase at the unit compared to other wards in the hospital. Secondly, some patients died rapidly, some other left against medical advice (LAMA) and the remaining discharged with medical advice without improvement due to small bed number of the unit. With this regard, if the patient stayed longer than other patients and any improvement was observed, the bed was left for new stroke patients. Contrary to this, few patients were stayed in hospital for greater than 3weeks. Multiple comorbidities, failed to show improvement and delayed in complimentary evaluations were some of the most feasible explanations for the prolonged length of hospital stay. This delayed evaluation will not only significantly increases the costs of stroke care, but also increases the risks of infection, other complications and recurrence in patients with suboptimal treatment and evaluation.
The in-hospital mortality of stroke patients (21.6%) was comparable with the study done by De Carvalho et al in Brazil 20.9%  and Desalu et al in Nigeria 23.8% . However, it was higher as compared to study done by Deresse et al in Ethiopia 14.7% , Tirschwell et al in Vietnam 6.5% , Masood etal in Pakistan 9 % , Gebremariam etal in Ethiopia 12.0 % , Greffie et al in Ethiopia 13.3%  and Jowi et al in kenya (5%) . But, the in hospital mortality rate in this was lower than the study done by Damasceno et al in Mozambique which was 33.3% , Atadzhanov etal in Zambia 40 %  and Walker et al in Gambia 57% . This difference could be due to different ways of stroke diagnosis, types of stroke, treatment approaches, comorbidities, complications and in hospital patient care.
The prominent immediate causes of death suspected by clinicians were increased intracranial pressure and respiratory failure secondary to aspiration pneumonia, which complies with other studies particularly conducted in Ethiopia [10, 21]. Furthermore, study done in Arabian Gulf countries reported that both neurologic and systemic complications accounted 63% of in hospital mortality . However, it was different from study conducted by Walker et al in Gambia in which the most immediate cause of death was the initial stroke itself in 61% of patients . The difference might be due to difference in physician’s assessment and prediction based on comorbidities as well as complications that were developed by the patients at the end of their life. Early identification and management of complications such as increased ICP and aspiration pneumonia will save life of some patients.
Generally, in hospital mortality rate of stroke in current study was higher than reports from western counties, but quite similar to SSA studies. This difference might reflect the limited access to hospital care, limited staffing, shortage of facilities for diagnosis, lack of necessary therapy and insufficient number of hospital beds for prolonged period of care in LMICs including Ethiopia. In addition to this, culturally some caregivers belief that patients should die at their home of origin where they spent most of their lives with family members around and caring for them. Absence of treatment with thrombolytic and the low frequency of treatment with antiplatelet agents for patients with ischemic stroke as well as lack of evaluation with neuroimaging with suboptimal care might be the other explanation for increased mortality of stroke patients in LMICs. Targeted interventions that reduce and control risk factors could substantially reduce the burden of stroke in SSA .
The median survival time for patients who died in hospital was 4.38 days which was shorter as compared to study by Walker etal 7.5 days , Greffie etal 6 days  and Damasceno et al 6 days . However it was relatively comparable with the study done by Deresse and Shaweno in Ethiopia reporting that the median survival time of the patients was 4.5 days after hospital admissions . The high mortality rate in this study during the first one-week (17.2%) might be due to acute complications developed among patients such as raised intracranial pressure and aspiration pneumonia.
Brain edema, urine incontinence, NIHSS>13 during hospital arrival and diagnosis of stroke clinically alone were the independent predictors of time to in hospital mortality. Except stroke severity, other factors were not reported on study conducted by Atadzhanov et al in Zambia . In this study increased NIHSS was associated with stroke severity constituting decreased level of consciousness. High NIHSS score as a predictor of mortality was consistent with previous study by Deresse and Shaweno in Ethiopia . However, according to study by Sweileh et al stroke subtype was one independent predictor of in-hospital mortality among stroke patients .
In current study, increased ICP (brain edema) was one predictor of in hospital mortality unlike study by Mamushet et al in Ethiopia in which mortality was not significantly associated with increased intracranial pressure . We believe that the number of in hospital complications were a reflection of the severity of stroke attack and it was an independent predictor of in-hospital mortality. The difference in predictors of in hospital mortality might be due to sample size, study design, significance value considered and eligibility criteria of the patient.
The in hospital mortality was higher for hemorrhagic stroke compared to ischemic stroke patients that complies with previous study findings [21, 22, 24]. Similar to study done by Das et al early onset mortality was common in hemorrhagic stroke, where late mortality was prevalent among ischemic stroke patients . In contrary to this finding, study by Mamushet etal showed that mortality was significantly higher for ischemic stroke compared to hemorrhagic stroke cases (P=0.049) . The difference might be due to the study design, study population, stroke subtype and comorbidity of the cases. Similar to our finding, study by Deresse et al showed that the rate of stroke related mortality was not different by age and sex .
The major strengths of this study was its prospective study design and the enrollment of consecutive patients which allowed us for collection of reliable data on time-varying prevalence of multiple variables. We used core and supplementary ascertainment strategies, combined with an independent direct assessment, to achieve recommended gold-standard findings. Inclusion of first-ever and recurrent stroke cases during the study period would provide a more accurate reflection of the burden of stroke.
The study provided a preliminary database on mortality and functional outcomes among stroke patients during discharge which can pave the way for stroke management strategies. The degree of the neurologic deficit on discharge was evaluated based on functional status score, unlike in most of other studies which was categorized into those with and those without neurologic deficit. We have also performed a detailed NIHSS assessment allowing us to evaluate for determinants of outcome in series of patients with stroke. In addition, we have used survival analysis method with competing risk that allowed us to estimate the risks of stroke mortality.
The study was associated with some limitations. First, this study was a hospital-based study rather than large community based study. Hospital based study may not reflect true picture of the stroke as extremely critical patients died before hospitalization and mild cases may have not reported to hospital. Additionally, hospital based study is subjected to referral bias, as most of the acute stroke patients’ visit our hospital only from the south western part of Ethiopia directly without any selection. These referral bias, single setup as well as convenience sampling approach used might not reflect the true burden and outcome of the stroke in our community. Hence, extrapolations and generalization to the rest of the community should be done with caution. Even though the study was hospital based, having only one referral center might probably reflect the actual magnitude of stroke in our country. As well as, the mean age, the proportion of young adults, male predominance, incidence and mortality indices of our data were quite similar to other stroke epidemiological studies.
Secondly, etiologic investigation for stroke was infrequently performed due to the lack of systematic cardiological examinations and brain imaging.it was evident in this study that about half of the patients were diagnosed clinically alone for stroke. Diagnostic investigations were undertaken on the basis of the subjective findings indicating inadequate workup and hence, possible underestimation. Finally, the sample size was small hampering the analysis of some prognostic indicators due to the short recruitment period and lack of resources. Indeed, a prospective community-based cohort design will be required thousands of stroke-free subjects who need to be followed up for several years to know the outcome of patients.