The authors have declared that no competing interests exist.
Conceived and designed the experiments: JAC VPM WS. Performed the experiments: JAC VPM WS HMB MPR. Analyzed the data: AMM EAR HMB. Contributed reagents/materials/analysis tools: JAC. Wrote the paper: AMM EAR HMB MPR VPM WS JAC.
In 2010, the World Health Organization (WHO) published updated guidelines emphasizing and expanding recommendations for a parasitological confirmation of malaria before treating with antimalarials. This study aimed to assess differences in historic (2007–2008) (cohort 1) and recent (2011–2012) (cohort 2) hospital cohorts in the diagnosis and treatment of febrile illness in a low malaria prevalence area of northern Tanzania.
We analyzed data from two prospective cohort studies that enrolled febrile adolescents and adults aged ≥13 years. All patients received quality-controlled aerobic blood cultures and malaria smears. We compared patients' discharge diagnoses, treatments, and outcomes to assess changes in the treatment of malaria and bacterial infections.
In total, 595 febrile inpatients were enrolled from two referral hospitals in Moshi, Tanzania. Laboratory-confirmed malaria was detected in 13 (3.2%) of 402 patients in cohort 1 and 1 (0.5%) of 193 patients in cohort 2 (p = 0.041). Antimalarials were prescribed to 201 (51.7%) of 389 smear-negative patients in cohort 1 and 97 (50.5%) of 192 smear-negative patients in cohort 2 (p = 0.794). Bacteremia was diagnosed from standard blood culture in 58 (14.5%) of 401 patients in cohort 1 compared to 18 (9.5%) of 190 patients in cohort 2 (p = 0.091). In cohort 1, 40 (69.0%) of 58 patients with a positive blood culture received antibacterials compared to 16 (88.9%) of 18 patients in cohort 2 (p = 0.094). In cohort 1, 43 (10.8%) of the 399 patients with known outcomes died during hospitalization compared with 12 (6.2%) deaths among 193 patients in cohort 2 (p = 0.073).
In a setting of low malaria transmission, a high proportion of smear-negative patients were diagnosed with malaria and treated with antimalarials despite updated WHO guidelines on malaria treatment. Improved laboratory diagnostics for non-malaria febrile illness might help to curb this practice.
Fever is a common symptom among adults seeking healthcare in sub-Saharan Africa
In 2010, the World Health Organization (WHO) updated its guidelines on malaria diagnosis, increasing the strength of the recommendation for parasitologic diagnosis of malaria across all age groups and in all levels of malaria transmission intensity. The reported aims of these changes included “prevention of unnecessary use of antimalarials” and “identification of parasite-negative patients in whom another diagnosis must be sought”
There are many possible reasons for misattribution of bacterial febrile illness as malaria
We studied the diagnosis and treatment of febrile illness in an area with low malaria endemicity by retrospectively examining diagnostic results and treatment decisions for two cohorts of adult inpatients
Moshi, Tanzania is a town of 144,000 inhabitants and the administrative center of the northern region of Kilimanjaro (pop. 1.37 million)
Kilimanjaro Christian Medical Centre (KCMC) is the referral hospital for Tanzania's northern zone and has 457 inpatient beds. KCMC serves a catchment area that includes the regions of Kilimanjaro, Arusha, Tanga, Manyara, and Singida. Mawenzi Regional Hospital (MRH) has 360 inpatient beds and serves as the regional hospital for Kilimanjaro.
Routine diagnostic tests to determine the cause of febrile illness were limited at both hospitals. Automated complete blood counts (CBC) with differential were available at both hospitals, as were gram stains of cerebrospinal or abscess fluid. Outside of the research study, both hospitals provided malaria smears; neither hospital offered malaria rapid diagnostic tests. No other rapid tests (i.e. dengue, cryptococcal antigen) were available at either hospital. Blood cultures were not available outside of the research study.
We analyzed data from two prospective cohort studies designed to elucidate the causes of fever among hospitalized patients in Moshi. Cohort 1 enrolled adolescents and adults aged ≥13 years from 17 September 2007 to 31 August 2008. Cohort 2 enrolled the same age group starting on 26 September 2011 at MRH and on 2 February 2012 at KCMC. Enrollment was ongoing during the preparation of this manuscript. We excluded cohort 2 patients enrolled before 3 October 2011 or after 27 September 2012 to maintain cohort lengths of similar duration. As a result, enrollment at KCMC occurred for approximately seven months (2 February 2012 to 27 September 2012) in cohort 2.
In cohort 1, patients with oral temperatures ≥38.0°C were eligible to participate. In cohort 2, patients with subjective fevers or temperatures ≥38.0° were invited to participate. In cohort 2, temperature was either tympanic (88.2%) or axillary (11.8%). For this analysis, we only included patients with measured temperatures ≥38.0°.
Trained clinical officers obtained a clinical history and physical examination on consenting patients using a standardized clinical review form. They also collected information on treatment received prior to admission and past HIV testing results. Health care providers from the two hospitals assigned patients a preliminary diagnosis before test results were available. This analysis does not take into account the results of specific tests performed by treating clinicians outside of the research study, including malaria smears that may have been performed as part of routine care at the participating hospitals' clinical laboratories. All malaria results reported herein were conducted in the study research laboratory (described below).
After skin cleansing with isopropyl alcohol and povidone iodine, blood was drawn for aerobic blood culture and malaria smear. In the remainder of this manuscript, we refer to positive and negative peripheral malaria smears as smear-positive and smear-negative. Additional laboratory evaluations were performed in cohort 1
All research samples were processed in the Kilimanjaro Clinical Research Institute (KCRI) Biotechnology Laboratory. The laboratory is known in the region for its high quality results and at the time of publication it is the only local laboratory which participates fully in international external quality assurance programs. These programs include College of American Pathologists and One World Accuracy, which include regular parasite surveys in order to meet standards for diagnosis of malaria films. All laboratory investigations were conducted according to Good Clinical Laboratory Practices standards.
Thick and thin blood films were stained with Giemsa and examined for blood parasites using oil immersion microscopy. Standard methods were used to determine parasite density
The results of cohort 1 analyses were published and heads of department (medicine and pediatrics at KCMC and chief of staff at MRH) were authors
Data from the standardized clinical review forms were entered into an Access database (Microsoft, Redmond, WA, USA) through the Cardiff Teleform system (Cardiff, Highland Park, IL, USA). We carried out a Kruskal-Wallis test to assess differences in continuous responses. Pearson's chi-squared test was used to compare categorical and binary responses. A multivariate logistic regression model, adjusted for applicable covariates, was used to look for associations with patient mortality. Statistical analyses were carried out using STATA version 12 (Stata Corp LP, College Station, TX, USA). All statistical tests were 2-sided and used probability values (p-values) of 0.05.
To determine which participants might require antibacterial treatment based on available data, three physicians (HB, MR, ER) were independently provided a list of all the discharge diagnoses from study participants. Those diagnoses deemed to require antibacterials by all three physicians were considered to have an expert recommendation for antibacterials. Based on this assessment, we deemed the following diagnoses as having an expert recommendation for antibacterials: abscess, adenitis, brucellosis, cellulitis, endocarditis, enteric fever, meningitis (excluding cryptococcal meningitis), pneumonia, septicemia, tuberculosis, and urinary tract infection. In addition, patients reporting symptoms which would prompt treatment with antibacterials according to the Integrated Management of Adult and Adolescent Illness (IMAI) were also considered to require treatment with antibacterials; these included patients with history of convulsions or stiff neck in the setting of fever
Because HIV testing was not performed for every patient in cohort 2, we carried out a sensitivity analysis adjusting for various levels of HIV prevalence and its effect on bacteremia prevalence. This analysis estimated the degree to which changes in baseline HIV infection might have contributed to differences in the prevalence of bacteremia. The adjusted proportion of patients with bacteremia was calculated assuming (1) that the HIV prevalence in cohort 1 was the same as the HIV prevalence at admission (i.e. that the only HIV-seropositive patients were those who tested positive prior to admission) and (2) that the prevalence was equal to the prevalence in cohort 1 (39.0%).
The two studies were independently approved by the KCMC Research Ethics Committee, the Tanzania National Institutes for Medical Research National Research Ethics Coordinating Committee, and the Duke University Medical Center Institutional Review Board. All participants, or their parent or guardian in the case of participants <18 years of age, provided written informed consent.
A total of 6353 patients were screened for enrollment in cohort 1 and 403 (6.3%) were enrolled; 3810 were screened for entry into cohort 2 and 340 (8.9%) were enrolled. For this analysis, we only included the 402 patients in cohort 1 and 193 patients in cohort 2 with measured temperatures ≥38.0°. The flow of enrollment for cohort 1 and 2 is outlined in
Demographic and basic clinical information for the two cohorts is shown in
n (%) |
n (%) |
||
Age, median (range) | 36.5 (13–95) | 37 (13–79) | 0.984 |
Female | 217/402 (54.0) | 120/193 (62.2) | 0.059 |
MRH admissions | 231/402 (57.5) | 138/193 (71.5) | |
Urban | 171/351 (48.7) | 92/170 (54.1) | 0.248 |
Rigors | 289/399 (72.4) | 115/151 (76.2) | 0.377 |
Headache | 284/398 (71.4) | 111/151 (73.5) | 0.616 |
Cough | 260/400 (65.0) | 111/193 (57.5) | 0.078 |
Vomiting | 140/400 (35.0) | 94/193 (48.7) | |
Shortness of Breath | 136/398 (34.2) | 60/193 (31.1) | 0.455 |
Fever >7 days | 85/399 (21.3) | 68/193 (35.2) | |
Diarrhea | 65/396 (16.4) | 33/193 (17.1) | 0.834 |
Stiff neck | 29/399 (7.3) | 21/151 (13.9) | |
Convulsions | 24/389 (6.2) | 7/193 (3.6) | |
Hemoptysis | 21/398 (5.3) | 5/151 (3,3) | 0.333 |
Jaundice | 7/396 (1.8) | 8/151 (5.3) | |
Past HIV test | 203/401 (50.6) | 121/193 (62.7) | |
Past HIV-seropositive test | 97/401 (24.2) | 51/191 (26.7) | 0.509 |
All confirmed HIV-seropositive | 157/402 (39.1) | N/A |
N/A |
Prior antimalarials | 174/396 (43.9) | 79/192 (41.1) | 0.521 |
Prior antibacterials | 170/398 (42.7) | 93/192 (48.4) | 0.190 |
Prior antiretroviral therapy |
53/97 (54.6) | 35/52 (67.3) | 0.134 |
Prior SXT prophylaxis |
52/96 (54.2) | 37/51 (72.5) |
Significant results are marked in bold.
* Denominators less than 402 (cohort 1) and 193 (cohort 2) represent missing values.
Questions on rigors, headache, stiff neck, hemoptysis, and jaundice were added mid-way through the study period.
Significance tests for comparisons between Cohort 1 and Cohort 2 determined by Kruskal-Wallis test for continuous variables and Pearson's chi-square test for categorical variables.
HIV testing was not routinely performed on patients in Cohort 2.
Among those with previous HIV+ test.
MRH: Mawenzi Regional Hospital; HIV: human immunodeficiency virus; SXT: trimethoprim-sulfamethoxazole.
In cohort 1, 203 (50.6%) of 401 patients had a previous HIV test versus 121 (62.7%) of 193 patients in cohort 2 (p = 0.006). Previous HIV-seropositivity was reported by 97 (24.2%) of 401 patients in cohort 1 and 51 (26.7%) of 191 in cohort 2 (p = 0.509). In cohort 1, 157 (39.1%) of 402 patients tested seropositive for HIV, 60 (38.2%) of which were among patients with no previous positive result. Because provider-initiated HIV testing had been adopted in Tanzania and all hospitalized patients were recommended to obtain HIV tests
In cohort 1, 53 (54.3%) of the 97 previously HIV-seropositive patients were receiving antiretroviral therapy (ART) compared to 35 (67.3%) of 52 HIV-seropositive patients in cohort 2 (p = 0.134). However, 52 (54.2%) of 96 HIV-seropositive patients in cohort 1 were receiving trimethoprim-sulfamethoxazole (SXT) prophylactic therapy versus 37 (72.5%) of 51 patients in cohort 2 (p = 0.030).
Malaria smear results for the two cohorts is shown in
n (%) |
n (%) |
||
Malaria smear positive | 13/402 (3.2) | 1/193 (0.5) | |
Adequate blood volume for culture | 365/401 (91.0) | 152/190 (80.0) | |
Bacterial culture positive | 58/401 (14.5) | 18/190 (9.5) | 0.091 |
Bacterial culture positive (adjusted %) |
13.8 | 9.8 | 0.194 |
Positive cultures arriving in time to influence clinical decisions |
43/58 (74.1) | 17/18 (94.4) | 0.065 |
Significant results are marked in bold.
* Denominators less than 403 (cohort 1) and 340 (cohort 2) represent missing values (except for culture arrival in time to influence clinical decisions).
Significance tests for comparisons between Cohort 1 and Cohort 2 determined by 2-sample t-test for continuous variables and Pearson's chi-square test for categorical variables.
Adjusted for adequate blood volume for culture, previous HIV testing, prior SXT prophylaxis, hospital location, and rurality.
Culture results received at least one day before patient discharge or death.
n (%) | n (%) | % | % | |||
Days in hospital median (range) | 5 (1–300) | 5 (1–44) | N/A | N/A | N/A | |
Malaria preliminary diagnosis |
150/402 (37.3) | 61/193 (31.6) | 0.173 | 37.3 | 27.6 | |
Malaria discharge diagnosis |
122/402 (30.3) | 45/193 (23.3) | 0.074 | 29.3 | 18.8 | |
Malaria smear-negative diagnosed with malaria |
110/389 (28.3) | 44/192 (22.9) | 0.168 | 27.4 | 18.7 | |
Malaria smear-negative treated with antimalarials |
201/389 (51.7) | 97/192 (50.5) | 0.794 | 53.5 | 46.4 | 0.132 |
Malaria smear-negative diagnosed with malaria given antibacterials |
50/110 (45.5) | 29/44 (65.9) | 45.2 | 66.5 | ||
Malaria smear-negative treated with antimalarials given antibacterials |
135/201 (67.2) | 79/97 (81.4) | 66.9 | 82.6 | ||
Preliminary diagnosis with expert recommendation for antibacterials |
131/402 (32.6) | 70/193 (36.3) | 0.374 | 32.6 | 36.2 | 0.389 |
Discharge diagnosis with expert recommendation for antibacterials |
108/402 (26.9) | 76/193 (39.4) | 27.0 | 38.8 | ||
Patients with indication for antibacterials |
164/402 (40.8) | 98/193 (50.8) | 39.1 | 48.4 | ||
Indication for antibacterials treated with antibacterials |
135/164 (82.3) | 94/98 (95.9) | 82.7 | 96.4 | ||
Bacteremic treated with antibacterials |
40/58 (69.0) | 16/18 (88.9) | 0.094 | 68.9 | 89.5 | 0.061 |
Antibacterial prescription for bacteremic patients with culture results arriving in time to influence clinical decisions | 31/40 (77.5) | 15/17 (88.2) | 0.347 | |||
Mortality |
43/399 (10.8) | 12/193 (6.2) | 0.073 | 7.4 | 5.6 | 0.371 |
Significant results are marked in bold.
* Significance tests for comparisons between cohorts determined by Kruskal-Wallis test for continuous variables and Pearson's chi-square test for categorical variables.
Adjusted for hospital location.
Presenting symptoms of stiff neck or convulsions, positive cultures arriving in time to influence clinical decisions or discharge diagnosis with strong indication for antibacterials.
Unable to calculate adjusted means because of small sample size.
Adjusted for hospital location and known HIV-serostatus.
Among smear-negative patients treated with antimalarials, antibacterials were prescribed to 135 (67.2%) of 201 patients in cohort 1 and 79 (81.4%) of 97 patients in cohort 2 (p = 0.010). Considering the two cohorts combined, antibacterials were prescribed to 214 (71.8%) of the 298 smear-negative patients treated with antimalarials compared to 231 (88.5%) of 261 smear-negative patients who did not receive antimalarials (p<0.001).
Blood culture results for the two cohorts are shown in
Information on the diagnosis and treatment of bacterial infections is shown in
In cohort 1, 164 (40.8%) of 402 patients and, in cohort 2, 98 (50.8%) of 193 patients had an indication for antibacterials (i.e. symptoms of stiff neck or convulsions, bacteremia arriving at least 24 hours before discharge or death, and discharge diagnosis with an expert recommendation for antibacterials) (p = 0.022). Among patients with a strong indication for antibacterials based on presenting symptoms, culture results, or discharge diagnoses, antibacterials were provided to 135 (82.3%) of 164 in cohort 1 and 94 (95.9%) of 98 in cohort 2 (p = 0.001).
Information on patient mortality in the two cohorts is shown in
Our results indicate that in an area of low malaria prevalence, febrile illness was often diagnosed and treated as malaria. Such management did not change from cohort 1 to cohort 2 despite increasingly greater emphasis from the WHO to limit antimalarial treatment to laboratory-confirmed malaria cases, low malaria prevalence throughout the study periods, and dissemination of information gleaned from cohort 1 to healthcare providers. Given our conservative definition of laboratory-confirmed malaria
One-tenth of febrile inpatients had positive blood cultures and bacteremia was seven-times more common than laboratory-confirmed malaria. Nevertheless, malaria was diagnosed as often as bacterial infections with an expert recommendation for antimicrobial treatment. In addition, while every smear-positive patient was treated with antimalarials, nearly 15% of patients with expert recommendations for antibacterial treatment did not receive antibacterials. However, more patients with a strong indication for antibacterials received them in cohort 2 compared with cohort 1. Among all patients, smear-negative patients treated with antimalarials were less likely to receive antibacterials. These findings correspond with earlier reports showing an association between use of antimalarials in smear-negative patients and non-treatment with antibacterials
Our study demonstrated a non-significant trend towards a decline in prevalence of bacteremia and mortality between the two cohorts. If HIV prevalence was indeed lower in cohort 2, this may have influenced such findings. In addition, an increased proportion of patients in cohort 2 reported use of antiretroviral medications and TMP/SMX prophylaxis. Alterations of other environmental or community risk factors or normal fluctuations in disease patterns may have contributed to this trend.
While enhancing local standards of care to include availability of blood culture and antimicrobial susceptibility would greatly improve the accurate diagnosis of local pathogens, other infectious diseases such as leptospirosis, rickettsioses, and arboviral infections were found to be common causes of febrile illness in cohort 1
This study had the strength of examining two relatively large cohorts of patients with febrile illness from the same two hospitals before and after changes in the WHO guidelines for malaria diagnosis and treatment
Taken together, these results indicate that updates in WHO guidelines and a consistent emphasis on parasitologic diagnosis have not been sufficient to change the practice of frequently diagnosing and treating “clinical” malaria among hospitalized adults in northern Tanzania. Laboratory results that are not available in time to affect clinical decisions, lack of confidence in results, pressure to conform to previous standards of care, inadequate dissemination of updated WHO guidelines for malaria treatment, and a lack of knowledge of other possible causes of febrile illness may all contribute to ongoing malaria overdiagnosis
We thank Ahaz T. Kulanga and Francis P. Karia, for providing administrative support to this study; Miriam Barabara, Michael Butoyi, Pilli M. Chambo, Beata V. Kyara, Beatus A. Massawe, Anna Mwalla, Anna D. Mtei, Godfrey S. Mushi, Lillian E. Ngowi, Flora M. Nkya, and Winfrida H. Shirima for reviewing and enrolling study participants; Gertrude I. Kessy, Janeth U. Kimaro, Euphrasia Mariki, Bona K. Shirima, and Edward Singo for managing participant follow-up; Jecinta J. Onyango for laboratory assistance; and Alphonce Mushi, Enock Kessy, and Evaline Ndosi for their roles in data entry. We acknowledge the Hubert-Yeargan Center for Global Health at Duke University for critical infrastructure support for the Kilimanjaro Christian Medical Centre-Duke University Collaboration. We are grateful to the leadership, clinicians, and patients of KCMC and MRH for their contributions to this research.