The present study used a methodological mix of quantitative, semi-quantitative and qualitative methods to determine the applicability of the revised dengue case classification system in 18 countries. Efforts were made to reduce observer bias and ensure robustness of data as discussed below.
Applicability of a dengue classification system in the clinical practice (objective 1)
The limited applicability of the DF/DHF Grade 1 and 2/DSS Grades 3 and 4 classification in clinical practice (particularly at PHC level), due to the rigidity of classification criteria and dependence on laboratory tests, has been highlighted in literature reviews . The prospective clinical DENCO study  further showed that in ~18% of dengue cases a correct application of the DHF criteria (fever, hemorrhagic tendencies, thrombocytopenia and plasma leakage expressed as raising haematocrit, pleural effusion and other signs and symptoms) was not possible. This has apparently lead to the clinical practice of diagnosing DHF without all the necessary criteria and the application of locally adapted variations . Our study highlights these issues. A large proportion - usually above 50% - of DHF criteria (such as haematocrit before and after treatment, platelet counts, and tourniquet tests), were not collected by the clinician; however, the expert reviewer of the medical charts used clinical judgement to come to the DHF diagnosis. But even doing so, 13.7% of patients could not be classified into the DF/DHF/DSS categories.
For the revised classification only 1.6% of dengue cases could not be classified and the warnings signs for severe disease (necessary for the "dengue +WS" category) were documented in a large proportion of medical charts (see the section on "completeness of information" in the results).
The revised classification also proved to be more sensitive for timely recognition of severe disease. Prospective chart reviews across all three levels of the health care system showed that a significant proportion of cases classified as DF by the DF/DHF 1 and 2/DSS 3 and 4 classification are picked up as potentially severe or severe (dengue plus warning signs or severe dengue) by the revised classification. In a direct comparison of the two classification systems, 51.9% (684/1327) of the DF cases were classified as dengue plus warning signs (revised classification) and another 5.7% (75/1327) as severe dengue (revised classification). Mismatches in the other direction, where cases were actually classified more severe by the DF/DHF 1 and 2/DSS 3 and 4 compared to the revised classification, were only present in a very small proportion, most of them being cases of DHF that were classified as "dengue without warning signs" according to the revised classification (N = 8, according to expert reviewer).
Applicability of a dengue classification system in dengue surveillance (retrospective chart review (objective 2)
The difficulties of applying the DF/DHF/DSS system in dengue surveillance are documented . In this study, the applicability of the two classification systems to retrospective data sets was tested. Here again the proportion of "non-classifiable" cases was higher when applying the DF/DHF/DSS system (12.5%) than when applying the revised case classification (only 3.1%). It is of particular concern that 32.1% of severe dengue cases could not be classified in the DF/DHF/DSS system. The completeness of information, particularly on potential warning signs, was lower in the retrospective chart review than in the prospective one - contributing to the 3.1% which could not be classified.
Usefulness of a severity classification for triage and case management (objective 1)
During an outbreak simple criteria for triage and case management are required to decide if patients can be treated at home, in a hospital ward or require intensive care. During inter-epidemic periods, endemic or sporadic transmission remains and triaging dengue cases for case management remains important for organizing health services. Our study is in line with the findings of the prospective DENCO study that DF, DHF and DSS only match to a limited degree with a more clinically oriented classification of severity. The revised classification was easily applicable in clinical practice (see above) but was also seen to be useful for triage and case management by medical staff, more frequently than the DF/DHF/DSS classification.
Yet, in the focus groups and questionnaires, some concerns on the revised classification were raised, these included:
hospitalization rates might increase if the warning signs are not precisely defined and therefore the need for a prospective study on the definition and usefulness of warning signs across different countries and health care levels was emphasized
cost implications if more patients are being admitted
the need for more training and dissemination and for more concise clinical protocols
The question if the revised classification leads to higher patient numbers cannot be answered with the results from this study and remains to be clarified. Issues related to triage and case management will have to be addressed in future studies.
The term prospective vs. retrospective refers to the timing of the training versus the treatment of the patients by the treating physician. Reviewers' assessment of the patient's classification was based on the medical chart after discharge of the patient. Thus, the knowledge of the outcome could theoretically have influenced the reviewer's assessment. However, as the reviewer had to check the presence of one or more items to classify severe dengue (revised classification) or a set of four items for DHF (current classification) the potential bias would most probably have worked in favour of the current classification in the sense that cases with severe clinical course would receive special attention to investigate the presence of all items needed to fulfil DHF in case of the old classification.
Acceptance and user-friendliness of a dengue classification system (objective 1)
Staff interviews and FGDs suggested that the level of acceptance for the revised classification was high. This was particularly the case for Latin America. Surprisingly it was also high in Asia, where the DF/DHF/DSS classification has a long tradition [15, 16].
The staff questionnaires also revealed that the revised classification was perceived as being more user-friendly compared to the DF/DHF/DSS classification, and that this has a direct implication on adequate case management and treatment. The respondents appreciated the revised classification with its emphasis on clinical severity which led directly to specific case management instructions in the clinical management algorithm. There was also an agreement across countries and through all methods used that the use of warning signs for case management, as reflected in the revised classification, is of practical value. However, it is obvious that both the definition of "warning signs for severe dengue" and the predictive value of these signs require further research. In addition, the question of a (clinical) dengue case definition in the absence of confirmatory laboratory tests needs further research. Taking into account the lessons learned from development of the revised classification, the development and dissemination of detailed guidelines on dengue clinical management is crucial - with emphasis placed on training for health staff at all levels. Furthermore, as suggested in a number of staff interviews, the recognition of special situations such as dengue in pregnancy, dengue and co-morbidities, dengue in paediatric and adult care need further consideration.
While there is a need for harmonizing guidelines (see ), it also may well be that algorithms in the dengue case definition may vary from region to region or even from country to country, and that some warning signs for severe dengue may show geographical variation. This will require the adaptation of dengue clinical guidelines to local characteristics. However, this study has shown that the revised dengue classification is suitable for clinical practice. For dengue surveillance, there is potential for the revised classification to lead to simpler, more consistent, and comparable data on dengue and severe dengue.