Nearly 20% of patients admitted to a Ugandan neurology ward died, with worse survival occurring among unemployed/retired persons and subsistence farmers/peasants and those with no diagnosis at time of death and a diagnosis of stroke. Among patients admitted to the neurology ward with a neurological diagnosis, there was a greater hazard of in-hospital mortality compared to admission to the same ward with a non-neurological diagnosis.
Several studies have analyzed the prevalence of neurologic illness within sub-Saharan Africa, many of which are community survey studies[4, 11, 18, 21, 30, 34]. This is the first report of the distribution of neurologic diagnoses and in-hospital mortality in a Ugandan neurology ward. Among community survey studies, one door-to-door survey study within a rural district of Uganda determined the point prevalence of neurological illness in the community was 3.3%, with peripheral neuropathy being most common (33.7%), followed by chronic headaches, stroke and epilepsy [18]. We reported a different distribution of neurological diagnoses, with differences likely attributable to the setting where sampling occurred (outpatient versus inpatient) and the geographic focus (one district in Uganda compared to every district in Uganda). The prevalence of neurological illness among hospitalized patients varies. One Nigerian inpatient study reported 24.2% of all inpatients had a neurological diagnosis [30], a study in an urban Ethiopian hospital reported a neurological diagnosis prevalence of 18% [4], and a study from Central Ghana reported 15% (of these, stroke comprised 54% and CNS infections 27%) [34]. Because our study included only patients admitted to a neurology ward, we do not report the prevalence of neurological illness amongst all patients presenting to Mulago Hospital.
Few studies in sub-Saharan Africa have reported the distribution of neurological diagnoses among inpatients with neurological illness admitted to a neurology ward and only one study reported predictors of mortality. One retrospective study from Cameroon found a similar mortality rate (19%) to that in our study with greatest mortality among those with stroke diagnoses (53%)[21]. Only one prospective study was identified in the literature reporting prevalence of neurological diagnoses in hospitalized patients in Congo, with a lower prevalence of stroke than that in our study (6.6% stroke) which may have been underestimated as there was no access to imaging [24]. Combined, these studies demonstrate that neurological illness is routinely encountered in community and inpatient settings throughout sub-Saharan Africa. Overall inpatient mortality in our study was comparable to other prevalence studies throughout sub-Saharan Africa from Ethiopia (22%) and Cameroon (19%) [4, 21], was lower than that reported by two studies from Nigeria (34%) and Central Ghana (31%) [30, 34], and was higher than one study from Congo (8.2%) [24] . Inaptient mortality due to neurological illness is variable, likely attributable to factors that differ between countries, including socioeconomic factors, healthcare systems, or data collection methods. The true mortality rate related to neurological illness in the country may have been underestimated, as data regarding patients that potentially died en route to the hospital or died in the Accident and Emergency ward prior to being hospitalized were not collected. In addition, patients were not included if they were admitted to a ward that was not the neurology ward. Therefore, patients who may have had a neurological illness and were inadvertently admitted to the medicine or another ward were unaccounted for, thus the hospital-wide prevalence of neurologic illness is unknown based on our study.
We did find that occupation was a strong predictor of poorer survival, such that patients in the lowest socioeconomic strata (i.e., subsistence farmer/peasant and unemployed/retired) had the highest in-patient mortality, compared with those who identified themselves as being employed or a farmer. Although not directly representative of socioeconomic status, there may be several factors that could be related to occupation as a surrogate for socioeconomic status that may have contributed to poor clinical outcomes, including poorer baseline health at the time of admission, longer distance to hospital if living in rural areas, poorer access to health care. This has been corroborated by other studies throughout SSA. One study in rural South Africa reported that lower socio-economic status was associated with higher HIV/AIDS, tuberculosis, and other communicable disease-associated mortality but no significant relationship between socioeconomic status and non-communicable diseases [17]. However, another study from the same region with an earlier time period 1994-2009 did show a similar, inverse relationship between socioeconomic status and mortality to that in our study [16]. Although these are plausible explanations, further work is needed to understand if these hypotheses explain this important finding.
Several other predictors of mortality were noted in our study, including an unknown admission diagnosis associated with increased mortality. Given limitations in diagnostic studies in our cohort, as in much of sub-Saharan Africa, more than a third of our cohort did not receive a diagnosis during their hospitalization. Diagnostic uncertainty is commonly encountered by health care providers in sub-Saharan Africa given the limited access to resources needed to evaluate patients. Research from several countries (South Africa [6], Mexico [15], China [31] and Tonga [7]) has identified substantial misclassification of in-hospital causes of death. While it is well-known that lacking a diagnosis leads to delay in appropriate care and subsequently increases the risk of death [9, 13], we are not familiar with other studies conducted within sub-Saharan Africa which report both the rate of unknown diagnoses during hospitalization and its influence on mortality. Though it stands to reason that interventions which provide resources necessary to conduct thorough evaluations would decrease the rate of unknown diagnoses and correspondingly improve mortality, this contention is speculative.
A diagnosis of stroke was predictive of poorer survival and was the most common diagnosis encountered in the Mulago Hospital neurology ward. The World Health Organization (WHO) reports that 85% of deaths globally attributable to neurologic conditions are due to cerebrovascular disease [8, 12, 39]. In sub-Saharan Africa, an estimated 9 to 13% of deaths are due to cardiovascular disease, including stroke [32], with yearly age-adjusted stroke rates four times higher in developing countries compared to developed countries [19]. Not surprisingly, recent work has sought to understand the true prevalence of stroke in sub-Saharan Africa. One community study in rural Uganda found stroke was one of the most common neurologic diagnoses with a prevalence of 14.3%, comparable to the worldwide stroke prevalence [18, 29]. Similarly, stroke prevalence among hospitalized patients in Ghana has increased from less than 2% in 1960 to 12% in 1993 [10]. Other studies throughout sub-Saharan Africa have found that mortality is high among stroke patients and is the leading cardiovascular cause of death and disability in sub-Saharan Africa, but further research is required to elucidate specific mortality predictors among stroke patients [23, 28, 33].
Reasons for increased mortality related to stroke may be explained by inadequate delivery of guideline concordant stroke care. For example, in the INTERSTROKE study, the mean time for completion of a CT or MRI of the brain during hospital stay was 30 hours, with vascular imaging having been performed in only 2.4% of the African cohort, transthoracic echocardiography was performed in a minority of patients (10%)[27]. Similarly, a Rwandan study reported median time to hospital presentation (from time last-seen-well to emergency room presentation) was 72 hours for ischemic stroke and no patients received thrombolytic therapy [26]. This is a stark contrast to the time to presentation in the U.S., usually within 6 hours [22]. In our study, no patient received thrombolytic therapy as neither the medication nor stroke protocols, which included rapid assessment of acute neurologic illness with urgent CT scan, were available. No MRI imaging was available in our study, thus the suspected diagnosis of ischemic stroke may have been under-represented in our study as diagnosis of stroke was assigned by the treating clinical provider based on clinical suspicion. Resources required for inpatient stroke management are unavailable or inaccessible to most patients given out-of-pocket costs patients must pay for testing; resources for secondary stroke prevention after discharge are often also inadequate.
While we identified several important predictors of mortality, including not receiving a diagnosis and receiving a diagnosis of neurological non-communicable diseases (e.g., stroke and head trauma), being diagnosed with a non-neurological non-communicable disease (e.g., diabetes, psychiatric illness) was not associated with in-hospital mortality in this study. Many of these non-neurological non-communicable diseases tend to be chronic conditions rather than disease processes that would increase short-term mortality, which may explain why patients admitted with a neurological condition (e.g., stroke), rather than a medical condition (e.g., hypertension), had higher rates of mortality. This is contrary to reports from the WHO that report 33% of mortality (including in-hospital and outpatient mortality) is due to non-communicable diseases, but this number includes non-communicable neurological diagnoses including stroke and head trauma and does not report in-hospital mortality [40]. In addition, these non-communicable diseases may lead to acute diagnoses such as stroke, sepsis, or other inpatient cause for admission, but the non-communicable disease itself may not have been considered the primary cause of death by the treating provider in our sample. Healthcare providers practicing in sub-Saharan Africa should be aware that patients admitted with neurological conditions to have higher in-hospital mortality compared to patients admitted with more chronic medical conditions.
Limitations to our study are worth noting. First, these data are subject to the known limitations of a cross-sectional study [3]. Second, this study was conducted entirely without the benefit of an EHR. EHR or a hospital-based patient registry would have allowed for a more comprehensive assessment of neurologic disorders across the hospital rather than limited to the neurology ward, thus our data on mortality rate and predictors of mortality is not generalizable across Mulago Hospital. While administrative data would likely allow for a more complete assessment of the relationship of patient-level factors and survival within the neurology ward of Mulago Hospital, our current work provides a sound description of the mortality rate and its associated predictors, including not being assigned a diagnosis by a treating physician, within a neurology ward. Until EHRs are more widely used in Uganda and other African countries, longitudinal data collection efforts similar to our own will be required to understand mortality and other outcomes among patients hospitalized with neurological illness as well as the impact of interventions designed to mitigate mortality. Third, our data collection was restricted to Ugandans admitted to the neurology ward with a presumed disorder severe enough to require inpatient treatment, and we did not collect data on death following hospital discharge. As mentioned previously, patients who died before being admitted from the Accident and Emergency ward or were admitted to the 4-bed intensive care unit would not have been accounted for. In our experience, it was not uncommon for patients to have a 2-3 day Accident and Emergency ward stay prior to arrival on the ward. Those with suspected CNS infections were often admitted to the infectious diseases ward and may have been under-represented. We have also found that many patients admitted with road traffic accidents (e.g., those involving boda bodas)[20] and severe head trauma who were unable to receive neurosurgical intervention from either of the two neurosurgeons, who covered the entire country of Uganda during the study period, experienced poorer outcomes prior to being hospitalized[35]. Also, patients admitted with a diagnosis of seizures may have been admitted to either the neurology ward or the psychiatric ward, dependent on the disposition of the treating provider. Given these considerations, our prevalence estimates may under-report the association between head trauma, seizures, infectious diseases and more fulminant disease and mortality. Fourth, treating providers may have provided a non-neurological diagnosis for the reason for admission or discharge (e.g., hypertension). While it is our experience that treating providers may have thought that non-neurological conditions were the reason for or contributed to a patient’s presentation, we did not formally interview healthcare providers their considerations in assigning specific diagnoses, or not assigning a diagnosis. Fifth, the neurologic diagnoses accounted for in this study were assigned by the treating clinical provider often without the benefit of neuroimaging, lumbar puncture and other diagnostic modalities routinely available in developed countries, with a minority of patients without an admission or discharge diagnosis. The treating clinical provider assigned a diagnosis based on their clinical judgement and any diagnostic tests available, however, we do not have evidence that the treating provider may have used a diagnosis from the past medical history when a primary neurological diagnosis was unknown. Given diagnostic uncertainty in resource-limited settings due to inaccessibility of many diagnostic modalities, diagnoses assigned were based on the medical judgment of the treating clinical provider. When the suspected diagnosis was unknown, no diagnosis was assigned or a known diagnosis from the past medical history (such as hypertension, atrial fibrillation, diabetes) thought to be contributing to the neurological condition was assigned when the neurological diagnosis was unknown. We did not identify reasons why providers assigned certain diagnoses to patients nor why certain providers assigned the patient to the neurology ward as the diagnosis assignment was left to the local Ugandan clinical provider who routinely cares for the neurology ward patients. However, these current data identify patients at increased risk of in-hospital mortality and can be used to guide quality improvement work directed at understanding more specific reasons related to mortality among these high-risk patients. Future work could also address reasons why certain diagnoses are assigned in these resource-limited settings. In addition, the study team did not follow patients after discharge. As such, we do not have prevalence and outcomes data of outpatient neurologic illness. Brain imaging, while potentially available to patients in-hospital, was obtained in a minority of patients. This was largely due to the inability of patients to pay out-of-pocket costs for these testing. Therefore, the diagnosis of stroke was largely made on clinical grounds when patients presented with such symptoms as sudden onset of paresis/paralysis, numbness, change in speech, or vision loss [14]. Because of this limitation, we analyzed ischemic and hemorrhagic stroke together, and could not report the prevalence of each stroke type. Finally, these data are from 2009 to 2011. More recent data collection would be required to understand the current state of neurological illness and mortality and allow for analysis of trends over time.
Despite these limitations, our study is one of the first to investigate the mortality rates and predictors of in-hospital morality on a neurology ward in sub-Saharan Africa. Future longitudinal work could focus investigations of health system factors that may be associated with in-hospital mortality (level of supervision and neurologic training of the clinical care team, access to an intensive care unit, healthcare worker strikes, refusal rates of lumbar puncture, availability and application of supportive care measures). By more thoroughly understanding the breadth and prevalence of neurological illness as well as predictors of poorer in-hospital survival, these data may serve to inform healthcare providers and policy makers about the development, implementation, and evaluation of interventions designed to mitigate mortality, especially among patients at high risk for poorer survival. Our findings specifically support further work on improving mortality among the Ugandans identified as being unemployed/retired persons and subsistence farmers/peasants, and those admitted with commonly encountered non-communicable (i.e., stroke, head trauma) and communicable disorders. These results also bring to light the rate and associated increased mortality of not having a diagnosis at admission or at time of death.