It is clear from the above discussion that large surveys need to include additional questions on diarrhea duration and severity to enable accurate measurement of treatment coverage. Such additional questions could use easily recognizable signs and symptoms, such as fever or abnormal thirst and vomiting to measure the prevalence of dehydration in the current or recalled episodes.
We identified two studies that showed various combinations of these signs as being predictive of dehydration 
. In a case control study in Brazil, researchers enrolled children who were hospitalized for dehydration as cases and age-matched children who also had diarrhea but were not hospitalized as controls. Mothers were asked to recount signs and symptoms on the first day of illness. The study concluded that using vomiting or fever as an indicator would identify 75% of diarrhea episodes with dehydration 
. In a similar study in Mozambique, vomiting or fever had a sensitivity of 68.3% (sensitivity measures how well an indicator is able to identify true positive cases, i.e., diarrhea with dehydration). By adding “drinking more than usual" to this combination, the sensitivity of diarrhea prediction increased to 87.8% with a specificity of 34.1% (specificity measures how well an indicator is able to identify true negative cases, i.e., diarrhea without dehydration) 
. We also found several studies that published risk factors for severe disease including socioeconomic factors, child characteristics, and some clinical signs and symptoms 
. Although additional research is clearly needed to refine and retest these signs and symptoms before introducing new questions in large-scale surveys across low- and middle-income countries, these data suggest that qualifying diarrhea severity at the community level is possible.
With the objective of demonstrating that the addition of simple questions to large-scale surveys might provide valuable insights into diarrhea severity and into treatment by diarrhea severity, we used the sensitivities and specificities provided by Victora et al. 
for combinations of reported signs and symptoms and 2×2 tables to calculate the positive and negative predictive values of these symptoms as predictors of dehydration during diarrhea. The positive predictive value (PPV) of a test indicates the proportion of individuals with a positive test result (here, diarrhea with dehydration) who actually have the disease being tested for. The negative predictive value (NPV) indicates the proportion of individuals with a negative test result who do not have the disease being tested for.
Victora et al. reported that the combination of “vomiting or fever or abnormal thirst" had the highest sensitivity (90%) and the lowest specificity (38%) () 
. The combination of “fever or vomiting" had the lowest reported sensitivity of 75% and the highest specificity of 66%. Typically, with any test, as sensitivity increases, specificity decreases and vice versa. Thus, as the signs or symptoms used to define diarrhea with dehydration become broader, more children meet the criteria of a “case" and the definition will capture a higher percentage of true cases. However, many more children who are not truly cases will also meet the case definition, which will increase false positivity and lower specificity.
Range of diagnostic values for dehydration from diarrhea based on selected severity indicators.
Unlike sensitivity and specificity, PPV increases as the disease prevalence increases. We therefore tested scenarios with a 5% and a 10% prevalence of dehydrating diarrhea. At these relatively low prevalences, the PPV for the different combinations of signs and symptoms did not vary widely (). Thus, for this set of specificities and sensitivities, additional questions about the presence of vomiting, fever, and abnormal thirst could correctly identify 75%–90% (i.e., sensitivity can be high) of diarrhea cases most in need of ORS (i.e., diarrhea with dehydration), but of those individuals that appear to have severe diarrhea using these signs and symptoms as a set of indicators, less than 20% will truly have an episode of diarrhea with dehydration (i.e., PPV is low).
Additional validation studies are needed to test these and other possible indicators in several settings before any disease severity questions are universally added to surveys. The specific wording of questions will also need to be studied across several locations and in several diverse cultures. Although other risk factors for diarrhea or severe diarrhea have been identified 
, we suggest that the focus in DHS and MICS questionnaires should remain on signs and symptoms of the episode that are simple to identify and recall. In addition to those reported by Victora et al. 
, questions on total days with diarrhea for completed episodes, number of days of illness for current episodes, and stools per day may further define the severity of the diarrhea episodes and improve our understanding of differences, if any, with regard to treatment or care seeking. These questions should be evaluated to determine if adding them to the survey will increase specificity and sensitivity of identifying cases of diarrhea at risk of progressing to dehydration.