Acute undifferentiated febrile illnesses in the tropics and sub-tropics are caused by a wide range of infectious diseases that often have indistinguishable clinical features. In developing countries there may also be insufficient microbiology facilities to identify these infections leading to missed diagnoses, inefficient use of healthcare resources, over-use of empirical treatments, a lack of information on antimicrobial resistance and inaccurate epidemiological data for guiding prevention strategies. These problems occur in Sri Lanka, but a prospective, systematic, representative and comprehensive study of febrile illnesses has never been performed. The Ragama Fever Study was performed at a major hospital in western Sri Lanka that served both urban and rural areas. Its aims were to identify the causes of febrile illnesses in a large sample of patients admitted to the hospital over a 1-year period, develop clinical prediction rules that could distinguish between the most common infectious diseases and assist in the evaluation of rapid (point-of-care) diagnostic tests that were appropriate to this setting. 617 (86.7%) of 711 febrile patients admitted to a quarter of the hospital medical wards were recruited. 56.4% had confirmed infections with organisms identified including dengue (22.2%), chikungunya (16.7%), leptospirosis (5.2%), various bacteraemias (4.2%), Q fever (2.9%), rickettsial infections (2.3%), tuberculosis (1.1%) and urinary tract infections (0.8%). 7.6% had confirmed infections with no organisms identified including cellulitis (2.4%), respiratory tract infections with radiographic changes (2.1%) and pulmonary tuberculosis with radiographic changes (1.6%). 4.1% had confirmed non-infectious diseases and 37.2% had unconfirmed diseases including “viral fever” (13.3%), undifferentiated fever (7.8%), respiratory tract infections (6.8%), urinary tract infections (3.4%), leptospirosis (2.8%) and gastroenteritis (1.0%). Clinical prediction rules for identifying dengue fever and chikungunya were developed using imputation, multiple logistic regression, scoring algorithms and receiver operating characteristic (ROC) curve analysis. The dengue fever rule had sensitivity = 49.6%, specificity = 93.9%, positive predictive value (PPV) = 70.8% and negative predictive value (NPV) = 86.1%. The chikungunya rule had sensitivity = 35.0%, specificity = 95.0%, PPV = 60.0% and NPV = 87.1%. ROC curve analysis could not identify any probability cut-offs that would produce clinical prediction rules with acceptable combinations of both sensitivity and specificity. A commercial (Panbio) rapid serology test for dengue fever showed sensitivity = 43.4%, specificity = 88.8%, PPV = 54.6% and NPV = 83.5% on samples from admission and significantly better diagnostic performance on follow-up. When repeated in conjunction with a PanBio rapid NS1 antigen detection test, the diagnostic performance improved with sensitivity = 89.9%, specificity = 75.0%, PPV = 69.0% and NPV = 92.3% on admission. This study confirmed the wide range of infections that present as febrile illnesses in Sri Lanka and showed the limitations of clinical prediction rules and rapid diagnostic tests in identifying these on admission. I hope that it will prove a foundation for further work on these important topics.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:579297 |
Date | January 2012 |
Creators | Bailey, Mark S. |
Contributors | Lalloo, David G.; de Silva, Janaka; Hart, Tony; Parry, Chris |
Publisher | University of Liverpool |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Source | http://livrepository.liverpool.ac.uk/9513/ |
Page generated in 0.0017 seconds