Objectives To identify case management, health system and antimalarial drug factors contributing to malaria deaths.
Method We investigated malaria-related deaths in South Africa's three malaria endemic provinces from January 2002 to July 2004. Data from healthcare facility records and a semi-structured interview with patients’ contacts were reviewed by an expert panel, which sought to reach consensus on factors contributing to the death. This included possible health system failures, adverse reactions to antimalarials, inappropriate medicine use and failing to respond to treatment.
Results Approximately 177 of 197 cases met inclusion criteria for the study. Delay in seeking formal health care was significantly longer for patients who sought traditional health care [median 4; inter-quartile range (IQR) 3–7 days] than for patients who did not (median 3; IQR 1–5 days; P = 0.033). Patients with confirmed or suspected HIV/AIDS were significantly more likely to use traditional approaches (25%) than those with other comorbidities (0%; P = 0.002). Malaria was neither suspected nor tested for at a primary care facility in 23% of cases with adequate records. Initial hospital assessment was considered inadequate in 74% of cases admitted to hospital and in-patient monitoring and management was adequate in only 27%. There were 32 suspected adverse reactions to antimalarial therapy.
Conclusion A confidential enquiry into malaria-related deaths is a useful tool for identifying preventable factors, health system failures and adverse events affecting malaria case management.
Objectifs Identifier les facteurs dans la prise en charge des cas, du système de santé et des médicaments antimalariques contribuant aux décès par malaria.
Méthode Nous avons étudié les décès liés à la malaria dans trois provinces endémiques pour la malaria en Afrique du Sud de janvier 2002 à juillet 2004. Les registres des services de santé et un entretien semi structuré avec les contacts des patients ont été passés en revue par une équipe d'experts qui a cherchéà atteindre un consensus sur les facteurs contribuant au décès. Cela comprenait: les échecs possibles du système de santé, les réactions adverses des médicaments antimalariques, l'utilisation inadéquate des médicaments et l'absence de réponse au traitement.
Résultats 177 sur 197 cas répondaient aux critères d'inclusion pour l’étude. Ceux qui ont fait recours à la médecine traditionnelle avaient un retard significativement plus long à recourir à la médecine conventionnelle (médiane 4 jours; Différence interquartile 3–7 jours) par rapport à ceux qui n'ont pas fait recours à la médecine traditionnelle (médiane 3 jours; Différence interquartile 1–5 jours; P = 0.033). Les patients avec VIH/SIDA confirmé ou suspectéétaient significativement plus probable de faire recours aux approches traditionnelles (9/36, 25%) que les ceux avec d'autres conditions de morbidité (0/32, 0%; P = 0.002). La malaria n'a vait pas été suspectée ou testée dans un service de santé primaire pour 22 (23%) sur 95 cas avec des données adéquates. L’évaluation initiale d'hôpital a été considérée insatisfaisante pour 125 sur 170 cas admis à l'hôpital (74%) et la prise en charge et surveillance des hospitalisés étaient adéquates chez seulement 46 sur 170 cas (27%). 32 réactions adverses suspectées au traitement ont été observés.
Conclusion Une enquête confidentielle sur les décès liés à la malaria est un outil utile pour identifier les facteurs évitables, les échecs du système de santé et les événements adverses affectant la prise en charge des cas de malaria.
Objetivos Identificar los factores que contribuyen a las muertes por malaria en el manejo de los casos, el sistema sanitario y los medicamentos antimaláricos.
Método Hemos investigado las muertes relacionadas con malaria en tres zonas de Sudáfrica en las que la malaria es endémica, entre Enero del 2002 y Julio del 2004. Un panel de expertos revisó las historias clínicas de los centros de salud, así como una entrevista semi-estructurada a los contactos de los pacientes, buscando consenso sobre los factores que contribuyeron en cada caso a la muerte. Entre estos factores se incluyeron posibles fallos sistémicos, reacciones adversas a los antimaláricos, el uso inapropiado de medicamentos y una respuesta fallida al tratamiento.
Resultados 177 de 197 casos cumplieron con los criterios de inclusión para el estudio. Aquellos que buscaban la ayuda de curanderos tradicionales presentaban un retraso significativamente más largo a la hora de buscar atención médica formal (media 4; IQR 3–7 días) que aquellos que no lo hacían (media 3; IQR 1–5 días; P = 0.033). Los pacientes con sospecha o confirmación de VIH/SIDA tenían más probabilidad de utilizar alternativas tradicionales (9/36, 25%) que aquellos con condiciones co-mórbidas (0/32, 0%; P = 0.002). En 22 de 95 casos con historias clínicas adecuadas (23%), no se sospechó o realizó ninguna prueba para malaria en centros de atención primaria. La evaluación hospitalaria inicial se consideró inadecuada en 125 de 170 casos admitidos en un hospital (74%) y la monitorización y el manejo durante el ingreso fue adecuado en solo 46 de 170 casos admitidos en un hospital (27%). Hubo 32 casos con sospecha de reacción adversa a la terapia antimalárica.
Conclusión El realizar un estudio confidencial de las muertes relacionadas con malaria es una herramienta útil para identificar factores prevenibles, fallos en los sistemas de salud y eventos adversos que afectan el manejo de casos de malaria.
Malaria continues to be a major cause of morbidity and mortality in Africa (Bremen et al. 2004). Approximately 1 million children and adults continue to die from malaria annually despite the increasing availability of effective preventive and curative interventions (Greenwood et al. 2005). It is widely acknowledged that operational research can be valuable in optimizing the benefits gained from wide-scale introduction of novel health technologies and disease control strategies (Durrheim et al. 2002; Harries 2004).
In South Africa, malaria transmission occurs in three provinces: Limpopo, KwaZulu Natal and Mpumalanga (Figure 1) (Malaria Research Programme. Medical Research Council 2005a). Low intensity seasonal transmission precludes the development of acquired immunity. Transmission is generally associated with higher rainfall and temperature, between September and May each year (Figure 2) (National Department of Health 2006). Plasmodium falciparum infection accounts for about 90% of malaria diagnosed in South Africa (Malaria Research Programme. Medical Research Council 2005b). After the introduction of more effective artemisinin-based combination therapy (ACT) in KwaZulu Natal in January 2001 (artemether lumefantrine; AL) and in Mpumalanga in January 2003 (artesunate plus sulphadoxine-pyrimethamine; AS + SP) and the strengthening of indoor residual spraying programmes, malaria notifications decreased by 84% (National Department of Health 2006; Barnes et al. 2005). These successes prompted the subsequent introduction of AL in Limpopo province in October 2004. However, annual provincial malaria case fatality rates have remained above 0.7% (range 0.1–1.5%) in South Africa (National Department of Health 2006). In October 2003, an outbreak of malaria occurred in a low-risk area overlapping Mpumalanga and Limpopo provinces, with a sharp increase in cases and deaths at that time.
Previous studies have shown that malaria deaths may serve as sentinel events, allowing determination of possible health system failures that could be addressed to prevent future deaths (Durrheim et al. 1999; Kain et al. 2001; Pfaff 2003; Newman et al. 2004). Detailed enquiry into each death provides an opportunity for elucidating common contributors to fatal outcomes. A case-fatality rate persisting above the national target of 0.5% and important differences between provinces (0.1–1.5%) prompted the current confidential enquiry. The purpose of the enquiry was to identify common contributory factors in case management and the health system to malaria-related deaths, with the intention of addressing these and thus improving the quality of patient care. In addition the introduction of new antimalarials in South Africa required confirmation of the safety profile of ACTs in public health programmes. The confidential enquiry was used to identify drug-related problems (such as adverse drug reactions, therapeutic errors and inappropriate drug use) and predisposing factors for adverse drug reactions.
Figure 3 summarizes the process by which included cases were collected, investigated and assessed. Malaria-related deaths are identified through the routine notification system and home-follow-up visits of notified malaria cases. The latter is routinely conducted by malaria control programme staff to confirm indoor residual insecticide spraying of the homestead of infected patients and to ensure that close contacts are tested for malaria. Deaths in the study period (January 2002 to July 2004 for Mpumalanga and KwaZulu Natal, and January 2003 to March 2004 in Limpopo) were investigated by the malaria programme staff. The notification system for malaria-related deaths was strengthened by training healthcare providers in notification of malaria and suspected adverse drug reactions.
In Limpopo, which has three times as many hospitals in the malaria region as the other two provinces, human resource constraints precluded investigation of all deaths. Here a stratified sample of deaths was selected to ensure that at least one death was investigated in each healthcare facility reporting any. Of 125 fatal cases in Limpopo province, 84 (67%) were investigated.
Each enquiry included a detailed review of all health records, whether public or private, and a home visit by the malaria control programme during which a semi-structured interview was conducted with the patient's contacts (i.e. family members, work colleagues or neighbours). All data were reviewed by an international expert panel (see Acknowledgements). Each case was reviewed at least three times; twice by individual panellists and the third time by the expert panel which met four times in November 2002, August 2003, March 2004 and November 2004. Repeated attempts were made to find missing data at each review stage, and these deaths were then reassessed at the next expert meeting. The panel sought to reach consensus regarding factors that may have contributed to the death, comprising health system failures and antimalarial adverse drug reactions, inappropriate antimalarial or concomitant medicine use and treatment failure. Adverse drug events were assessed to determine causal association according to the World Health Organization criteria for causality assessment (Edwards and Biriell 1994). The adequacy of clinical care was assessed against recommendations for rural district hospitals in the South African National Guidelines for the Management of Severe Malaria (Anonymous 2002), derived from the WHO treatment guidelines (Table 1) (Anonymous 2000). The panel was asked to propose feasible interventions to prevent similar adverse outcomes in the future.
Table 1. Criteria for adequacy of clinical care for patients with severe malaria
*Not recording only one vital sign and/or absence of a rapid test (when the patient had a confirmed test at the clinic) was still considered an adequate assessment.
**Monitoring and management of complications were considered inadequate if one or more of the following criteria were not met.
Initial admission note considered adequate if record of*:
Vital signs (heart rate, temperature, blood pressure and respiratory rate)
A random blood glucose level
Rapid test or malaria smear result
Neurological assessment or mention of the patient's mental status
A written order for haematocrit or haemoglobin, urea and creatinine and malaria smear on the admission note
Ongoing monitoring considered adequate if record of**:
Doctor's notes: At least one progress note entry per day from the day of admission to day of death, excluding the final death entry. Where the patient died before 7 am, a doctor's progress note was not considered necessary for that day
Nursing notes: At least one entry per 12 h shift (i.e. 7 am to 7 pm and 7 pm to 7 am) except for the day of admission and day of death if the death occurred before 7 am
Glucose: Presence of a glucose monitoring chart with regular entries (at least daily or more frequently as ordered in the doctor's notes)
Fluid status: Presence of an input/output chart with entries at least on 1 day of hospital stay, unless the patient was diagnosed with renal failure after which daily monitoring is required
Observation of vital signs: Presence of a temperature and BP observations chart in which at least daily blood pressure, temperature and pulse had been entered throughout the patient's hospital stay
Haemoglobin & platelets: At least one record of haemoglobin and platelet results during hospital stay
Urea & creatinine: At least one record of urea and creatinine
Laboratory services: Laboratory results reported either telephonically or in printed form within 8 h of receipt of samples
Deaths were included if they were considered associated with malaria, which was confirmed by rapid diagnostic test or malaria smear or its complications. Patients treated for confirmed malaria within 7 days of death, regardless of the stated cause of death, were also included in the enquiry. It was not possible to establish malaria as the definite cause of death for all cases, as relevant information on other causes may not have been requested or available. Cases were excluded: if (i) objective data confirmed a cause of death other than malaria, and no concurrent episode of malaria in the patient could be documented or if (ii) malaria was clinically suspected but never confirmed by rapid diagnostic test or malaria smear.
Data entry and analysis were performed using Microsoft® Excel 2003, Statistical Program for Social Sciences version 13.0 (SPSS, Inc., 2004) for Windows and EPI-Info version 3.01 (Centres for Disease Control and Prevention, Atlanta, GA, USA, 2003). Proportions were compared using χ2 with Yates correction or Fisher's exact test. Non-parametric data was summarized using median and inter-quartile range (IQR) and compared using the Mann–Whitney U or Kruskal–Wallis tests. Significance was determined at the 95% level.
Approval was granted by the three provincial Departments of Health and the University of Cape Town's Research Ethics Committee. Confidentiality of all patient records, care providers and health facilities was maintained.
During the study period a total of 29 329 malaria cases were notified of which 248 were fatal (0.8%). Among malaria cases notified from Limpopo, KwaZulu Natal and Mpumalanga provinces between 1 January 2002 and 31 July 2004, 55% were male. The majority (74%) of malaria cases occurred in adults, 15 years or older. Case fatality was significantly higher in adults compared with children under 15 (P = 0.003), and even higher in the elderly (>60 years) (P < 0.0001) (Table 2).
Table 2. Case fatality ratios (%) by age group showing number of malaria-associated deaths and malaria case notifications between 1 January 2002 and July 2004 (excluding four deaths in which age is unknown)
Case fatality ratios (malaria deaths/cases)
Odds ratios compared with 0–4 years CFR (95% CI)
*χ2 trend P < 0.0001.
0.31% (8/2 566)
0.32% (16/5 042)
P = 0.9674
0.87% (179/20 610)
P = 0.003
3.7% (41/1 111)
P < 0.0001
Of the 248 deaths notified during the study period, 197 were investigated (Mpumalanga 70; Limpopo 84; KwaZulu Natal 43). Twenty were withdrawn, either because malaria was excluded and another disease was confirmed as the cause of death (HIV-associated cryptococcal meningitis n = 3; HIV-associated TB n = 2), or malaria was never confirmed (n = 15). Healthcare records and interviews with contacts were available for 173 (98%) and 161 (91%) of the 177 cases respectively. Demographic details of the 177 included cases are provided in Table 3. In 50/177 (28%) cases, care was initially sought at a health facility in a low risk transmission area or where malaria transmission is not reported to occur. Twelve cases (7%) were reported to be foreign nationals. The median age of malaria-related deaths was 37 years (IQR, 24–49), excluding three adults with unknown age. The vast majority of malaria-related deaths were adults (220/248, 89%).
Table 3. Demographics of patients included in the confidential enquiry
n = 177 (%)
Patient's residential area malaria risk
Moderate risk area
Low or no risk area
Comorbidities were suspected or confirmed by health workers in 68/177 (38%) cases, of which 36/68 (53%) were thought to have HIV/AIDS. Fifteen (42%) of these 36 cases were confirmed as HIV positive by laboratory testing, while the remaining cases were diagnosed with AIDS-defining illnesses or on the basis of doctor's notes recording suspected HIV infection without documentation of any specific testing. Pregnancy was confirmed in six cases. Of 86 women for whom age was known, 50 (58%) were of child-bearing age (15–45 years).
Initial treatment seeking
Time until treatment seeking, as reported by patients’ contacts was a median of 3 days (IQR 1.5–5; range 0–14), excluding 57/177 (32%) cases for whom the duration of symptoms could not be accurately established. A delay of > 5 days was recorded in 23/120 (19%). Traditional beliefs and the assumption that the disease was self limiting were the most frequent explanations provided by relatives, neighbours or fellow employees for delayed treatment seeking (Table 4). Delay in seeking formal health care was significantly longer for those who sought traditional health care (median 4; IQR 3–7 days) than those who did not (median 3; IQR 1–5 days; P = 0.033). Fewer contacts of patients with confirmed or suspected HIV (19/36, 53%) could remember when malaria symptoms began than those with comorbidities (32/32, 100%; P = 0.029) or without comorbidities (77/109; 71%, P = 0.05). Patients with confirmed or suspected HIV/AIDS were significantly more likely to use traditional approaches (9/36, 25%) than patients with other concomitant illnesses (0/32, 0%; P = 0.002).
Table 4. Reason(s) given by relatives or colleagues for any perceived delay in seeking care
Reason for delay
Traditional beliefs or initial use of traditional medicine or approaches
Assumption that disease was self-limiting
Transport, financial or social constraints delayed access to care
Assumption that symptoms were related to a co-morbid condition (HIV-related condition)
Patient refused to visit a health care facility – no specific reason given
Reticence by employer to allow treatment-seeking or fear of dismissal by employee
Work or school responsibilities delayed treatment seeking
Health facility closed after hours
Primary health care (PHC) level (public clinic nurse or private general practitioner)
Sixty-six per cent sought care at a primary health service, such as a public sector clinic or hospital outpatient clinic, during the course of their illness, of whom 22 (12%) had inadequate records for evaluation. Malaria was not suspected nor tested for in 22 of 95 (23%) cases with adequate records at PHC facilities. Private sector providers were more likely to miss the diagnosis in fatal malaria cases compared with public sector PHC services (8/17, 47% vs.14/78, 17.9%; P = 0.02) based on the diagnosis subsequently made in hospital. Among those diagnosed at primary health level, disease severity was under-assessed in 10 of 73 (14%) cases.
Of the 177 malaria-related deaths, 170 (96%) were admitted to hospital, of whom 167 (94%) died while in hospital. The median hospital stay was 3 days (range 1–16). In 13/170 cases (8%) malaria was not tested on admission despite a history of typical symptoms. Facilities for confirming malaria were available at all study hospitals.
Initial assessment of disease severity at the time of admission was considered inadequate for a rural district hospital (as defined in Table 1) in the majority of cases (125 of 170, 74%). From available records of the 170 patients admitted, neurological status was only recorded in 99 (58%); glucose levels were recorded or requested in only 16 (9%), urea and creatinine levels in 87 (51%); full blood counts in 112 cases (66%) and malaria smears in 59 cases (35%). Twenty-three of the 59 cases (39%) in whom a parasite count was recorded had documented hyperparasitaemia (> 5% or +++ or parasite density ≥ 250 000 asexual parasites/mm3). A pregnancy test was only performed in 3/43 (7%) women of childbearing age who were not previously confirmed to be pregnant.
Based on criteria specified in Table 1, inpatient monitoring and supportive management was considered adequate in only 46 of 170 (27%) cases. Table 5 describes the nature and frequency of malaria complications recorded in the hospitalized patients. The most frequently recognized complications were neurological (84%). The low proportion of patients adequately investigated is likely to result in underestimation of the frequency of these complications. Complications were inadequately managed with limited input-output monitoring and poor fluid management being the predominant problem. Of 133 patients managed in hospitals for more than 24 h, input-output monitoring and fluid management was only adequate in 44 (33.1%).
Table 5. Frequency of malaria complications among 170 patients admitted to hospital
Number of patients/number assessed (% of cases assessed)
Disturbed consciousness or seizures (i.e. confusion, coma, restlessness or agitation)
Renal failure (serum creatinine >265 μmol/l)
Respiratory distress/pulmonary oedema (i.e. frothing, dyspnoea, tachypnoea or nasal flaring)
Jaundice and/or serum total bilirubin >43 μmol/l Note: Jaundice was clinically diagnosed in 32 patients
Hyperparasitaemia (>5% or +++ or parasite density ≥ 250 000/μl)
Drug-related problems and adverse drug reactions
A quinine loading dose was not administered in 36% (61/166) of patients admitted to hospital with severe malaria. In four cases, a second loading dose of quinine was administered following clinical deterioration of patients who had already received maintenance doses for at least 24 h and who had received a loading dose on admission. Six patients were inappropriately treated at hospital with oral antimalarials alone. In five patients records suggest that quinine was administered very rapidly, which may have contributed to the sudden death of one patient shortly thereafter. Of the 125 patients for whom hospital prescription charts were available, 37 (30%) did not receive all of their quinine doses as prescribed. Stock-outs of quinine were documented in 2/31 hospitals and expired quinine was found in one hospital.
There were 32 suspected adverse reactions for which a causal association with antimalarial therapy could not be excluded. Severe hypoglycaemia (glucose < 2.2 mmol/l) developed or was exacerbated after quinine treatment in 23 patients. Symptoms associated with cinchonism, particularly impaired hearing, were reported in five patients, and bradycardia in one. Three patients deteriorated clinically after administration of artesunate plus SP, but two of these were considered unlikely to be causal. Respiratory distress and hypotension developed in a 31-year-old man with repeated vomiting, who was inappropriately treated as an outpatient with artesunate plus SP for severe malaria. Severe haematuria developed in a 24-year-old woman who had been treated 13 days previously with artesunate plus SP for apparently uncomplicated malaria.
A 42-year-old man presented at a clinic with body pains, rigours and headache, with a blood pressure of 90/60 and was treated with AS/SP after a positive rapid malaria test. Although he had improved at follow-up after 2 days, 4 days later he developed dizziness, fell and injured his chin. At the rural district hospital his BP was 80/50 and temperature was 36.3 °C. An intramuscular stat dose of an unidentified medicine, lactated ringers infusion and a quinine 600 mg intravenous infusion were administered after another positive rapid malaria test. The blood glucose was 4.9 mmol/l. After a single dose of quinine he was discharged the next day on another course of AS/SP, with amoxicillin and paracetamol. His family reported that he became confused 2 days later, with a headache, difficulty in talking and breathing. He fell down that night and died at home. The father stated that his son had been weak for some time and attributed his death to this weakness. A causal link with artesunate plus SP treatment was possible in the light of the temporal associations, although information on his underlying chronic weakness was inadequate to exclude alternative diagnoses.
Non-response to treatment was noted in seven patients (three cases with AL, two of the above-mentioned adverse event cases with AS/SP and two cases with SP). A case was classified as a non-response to treatment if the patient failed to improve clinically or parasitologically despite completing an antimalarial treatment regimen. Using this retrospective approach, true antimalarial resistance could not be differentiated from non-response due to vomiting, undisclosed poor adherence or inappropriate oral treatment. A co-morbid disease may also have masked response to treatment.
The panel also identified possible problems with non-antimalarial drugs in 46/170 cases (28.9%). The most common included corticosteroids in nine patients despite an established association of corticosteroids with adverse outcomes in severe malaria (World Health Organization 2000); inappropriate use of benzodiazepines or other sedatives in restless, confused or hypoxic patients (eight cases), and failure to appropriately administer an antiepileptic (seven cases). Antibiotics were not prescribed when indicated in three cases and were administered orally when parenteral therapy was clearly warranted in one patient. Aminoglycosides (n = 4) and non-steroidal anti-inflammatory drugs (n = 4) were used in eight patients where there was a clear risk of renal failure.
Referral and patient transfer
In 15 patients, no attempt was made to transfer the patient to a referral hospital or high care unit despite clear indications for intensive care, ventilation or dialysis from hospital records. In a further 18 patients, attempts at referral were unsuccessful or delayed as transport was not available, the patient was too ill to be transported or the referral facility did not accept the patient.
Malaria deaths help to identify strengths and weaknesses in the interaction between patients and the health system (Durrheim et al. 1999; Kain et al. 2001; Pfaff 2003; Newman et al. 2004). This confidential enquiry successfully identified access to care issues, therapeutic problems and possible adverse effects from antimalarials. The study also identified additional malaria deaths not identified by the routine notification system and found deaths misclassified as being malaria-related. The multidisciplinary approach with critical review of individual records by an expert panel, against standard criteria for assessing the adequacy of case management, allowed a more thorough assessment than verbal autopsy alone (Todd et al. 1994). Most importantly, the three provincial Departments of Health responded to this locally gathered evidence by identifying and prioritising appropriate corrective measures for improving malaria management.
The value of confidential enquiries in improving disease management has long been recognized in developed countries (Cook 1989; Wareham et al. 1993; Papworth & Cartlidge 2005). The success of this method in the South African setting can partially be ascribed to reasonable access to health facilities with good diagnostic capability and the relatively low malaria case load compared with other African settings. In many other parts of Africa, the diagnosis of malaria-related deaths relies largely on verbal autopsies as a substantial proportion of deaths occur at home (Iley 2006). The method is limited by the potential misclassification of deaths occurring outside the formal health sector (Garenne and Feauveau 2006). This misclassification could dilute the impact of such assessments by including those cases incorrectly diagnosed as malaria, particularly in areas with a high prevalence of HIV (Makani et al. 2003). In Sub-Saharan Africa, confidential enquiries could be used to enhance verbal autopsy data and identify key health system problems contributing to high case fatality rates among patients seeking care in the formal health sector.
Although the CFR remained below 1% for the period of study, potentially correctable factors were identified at community, clinic and hospital levels. CFRs as low as 0.19% are possible in low transmission areas, as demonstrated in Thailand (Luxemburger et al. 1997). These differences in case fatality rates may be due to higher awareness in a small, confined population with better access to care and less comorbidities such as HIV/AIDS and TB.
Less than half of fatal cases sought formal health care within 48 h of symptom onset. Family members may have under-reported delays due to problems with recall, guilt and fear. Traditional treatments, either at home or from a traditional healer, led to delays in seeking medical care, particularly in cases where HIV/AIDS was suspected or confirmed. There is limited access to antiretroviral therapy in rural South Africa, and recurrent and persistent opportunistic infections may be confused with malaria. Community studies in South Africa when AIDS was less prevalent found that when malaria was suspected, patients would rapidly attend a formal health facility as they were aware of the dangers of malaria and knew effective treatment was available (Govere et al. 2000; Barnes et al. 2005). These studies included a broader population than this HIV-infected sub-group of fatal malaria cases.
Barriers to prompt treatment-seeking were similar to those recognized in other developing countries (Davies and Sophat 2004). Reasons for delay despite severe illness included parents sending children to school to write examinations, fear of dismissal, illegal immigrants’ fear of prosecution or deportation and lack of transport to a healthcare facility. These reflect the disproportionate effect of malaria on poor and vulnerable individuals (Amexo et al. 2004; Bremen et al. 2004).
In 2003 the HIV seroprevalence in the general South African adult population was estimated at 19–25%, one of the highest in the world (Anonymous 2004a). In this study 33 of 133 (25%) similarly aged adults had confirmed or suspected HIV/AIDS. The impact of HIV/AIDS co-morbid disease on malaria case management is the main difference between our study and confidential enquiries into malaria deaths previously conducted in South Africa (Durrheim et al. 1999; Pfaff 2003). In almost half the patients with suspected or confirmed HIV/AIDS the date of onset of disease was unknown. Patients were often chronically ill or being treated in the informal health sector. On admission to hospital, cryptococcal and TB meningitis were incorrectly diagnosed as cerebral malaria and it is not known what percentage of patients whose deaths were attributed to cryptococcal or TB meningitis may have actually died of malaria. Therefore, in addition to the documented biological interactions between malaria and HIV/AIDS, greater risk of malaria, higher parasite density and poorer response to standard antimalarial treatment, our findings suggest that where HIV/AIDS is prevalent, patients and physicians may not be able to differentiate between malaria and conditions related to HIV, with detrimental effects on malaria recognition and prompt treatment (Anonymous 2004b). Makani et al. found that cerebral malaria was grossly overdiagnosed in stable endemic areas with a relatively high prevalence of HIV/AIDS (Makani et al. 2003). Therapeutic management of cerebral complications, anaemia, fluid management and co-infections are likely to require more intense clinical and laboratory investigation and management in HIV co-infected patients. There should be a low threshold for conducting malaria testing in HIV-infected patients who present with fever in malaria-endemic regions.
Fatal malaria cases were generally well managed initially at public sector clinics with few missed diagnoses, few stock-outs of rapid tests, no stock-outs of drugs and appropriate referral to hospital. This may be due to ongoing staff training and reliable supplies of tests and drugs (Pfaff 2003). However, there was occasional inappropriate use of oral antimalarials in severely ill or vomiting patients and inadequate assessment for hypoglycaemia, despite a high number of altered level of consciousness and seizures. These issues need to be addressed in updated treatment guidelines. For remote clinics where delays in patient transfer are likely, providing intramuscular quinine could be considered.
Missed diagnosis and underuse of rapid tests were common problems in private medical practices. In almost half the fatal cases (8/17) who initially presented to a private practice, malaria was not suspected. The apparent lower level of clinical suspicion and limited availability of rapid diagnostic tests and effective treatments should be addressed through improved training and supervision of the private sector. The misdiagnosis of malaria can contribute to mistrust in allopathic medicine, delayed treatment seeking and ultimately increasing ill-health and deepening poverty, as seen elsewhere in Africa (Amexo et al. 2004, Makani et al. 2003).
Under-assessment of disease severity on admission was a common failing. Medical records were often sparse or incomplete, and relevant laboratory tests were not always requested or acted upon, despite being available at all hospitals. Incomplete assessment and monitoring, delayed recognition and management of complications were significant problems. These may reflect inexperienced medical staff and limited resources to manage complicated malaria. In addition to the initial diagnosis with rapid tests malaria smears were performed in a third of the investigated deaths. This precluded the detection of hyperparasitaemia, an important prognostic indicator of therapeutic efficacy and survival; even in cases with uncomplicated hyperparasitaemia, mortality triples (Price et al. 1998).
At a rural district hospital it is difficult to manage severe malaria complicated by renal failure and respiratory distress. Severe malaria has a mortality rate of 12–22% and should be managed at the highest level of care available (Newton et al. 2003). Thus, effective ambulance services and high care facilities should ideally be available in malaria transmission areas (Newman et al. 2004).
The expert panel suggested several factors which may have contributed to poor hospital care in general: high staff turnover, poor morale due to overwork and the growing burden of HIV/AIDS, inadequate supervision and training of junior staff and lack of routine handover by admitting staff to ward staff.
The few adverse reactions observed are consistent with the remarkable safety profile reported for the artemisinin based combination therapies (Taylor and White 2004). Of the suspected adverse drug reactions identified, the majority were related to inappropriate use of drugs. Failure to administer a quinine loading dose and the repeated or rapid administration of quinine loading doses following clinical deterioration in patients already receiving maintenance quinine therapy were major problems identified. The latter could result in toxic quinine levels which may induce fatal arrhythmias, particularly in patients with renal impairment. The inadequate monitoring of glucose levels is particularly problematic for patients on IV quinine. Recent evidence suggests that intravenous artesunate reduces mortality from severe malaria, and decreases the risk of hypoglycaemia and other adverse effects compared with IV quinine (South East Asian Quinine Artesunate Malaria Trial (SEAQUAMAT) group 2005). Our findings support the need for this safer, simpler and more effective alternative for treating severe malaria.
Incorrect drug use was not limited to antimalarials. Inappropriate use of benzodiazepines, corticosteroids and potentially nephrotoxic drugs were seen in severe malaria. These problems highlight the need for retraining of staff. All hospitals should conduct regular mortality review meetings to improve patient care.
We believe that the majority of malaria deaths were captured by this study, although some may have occurred outside the formal health sector or were missed diagnoses. In South Africa burial is only permissible on presentation of a death certificate signed by a medical officer. Formal burials are regarded as culturally important, enhancing compliance with this legal requirement. The cases included in this study, by definition, represent the worst outcomes and therefore may not represent the quality of care generally provided to patients with severe malaria. To assess general care and identify factors that increase the risk of malaria-related deaths, a case–control study of patients who survive severe malaria compared with those who did not survive, would be needed (Ninis et al. 2005). Problems identified in this study including delayed access to effective treatment, inadequate patient care, and inappropriate and ineffective drug treatment, have been shown to affect patient outcome in other studies (Trape et al. 1998; Taylor & White 2004).
Strengthening health districts and adequate malaria case management are vital even where new, more effective treatments have been introduced. Pharmacovigilance of new therapies in the normal context of use is also urgently needed. Encouraging practical, evidence-based treatment guidelines, supported by regular relevant training, efficient laboratory services, on-site supervision and simple patient monitoring tools, with feedback from routine clinical audits have demonstrated benefits for patient management (Durrheim et al. 1999; Pfaff 2003; Wahlstrom et al. 2003; English et al. 2004). HIV/AIDS co-infection is now a critical modifier of treatment-seeking behaviour, correct diagnosis and appropriate case management. A confidential enquiry into malaria-related deaths has proven a simple and effective tool for identifying preventable factors, health system failures and adverse events that may detrimentally affect the outcome of patients with malaria.
We thank the Mpumalanga, KwaZulu Natal and Limpopo Provincial Departments of Health for appreciating the value of this confidential enquiry and authorising healthcare facility records to be made available. The authors gratefully acknowledge the malaria control programme case investigators who collected all the healthcare facility records and conducted interviews with the next of kin. The cooperation of the healthcare facility staff and next of kin in providing the necessary data is appreciated. We acknowledge members of the expert review panel for their valuable contributions to ensuring the validity of this enquiry. The expert review panel members comprised: Prof. Karen Barnes, Dr Mark Blaylock, Dr Lucille Blumberg, Dr Armando Sanchez Canal, Dr Steven Donohue, Prof. David Durrheim, Dr Normal Cecilia Garcia, Mr J.K. Gumede, Dr Frank Hansford, Dr Bernice Harris, Dr Jakobus Hugo, Dr Etienne Immelman, Dr Melanie-Anne John, Mr Philip Kruger, Mr Aaron Mabuza, Dr Bonnie Maloba, Dr Ida Makwetla, Dr David Masipa, Dr Henry Okonta, Dr Isabela Ribiero and Dr Gerhard Swart. Special thanks go to Mrs Tracey Fourie for organizing expert panel review meetings. The South East African Combination Antimalarial Therapy (SEACAT) evaluation, within which this study was nested, received partial financial support from the UNDP/World Bank/WHO Special Programme for Research and Training in Tropical Diseases (TDR).