Kim Mulholland

To describe the epidemiology of respiratory syncytial virus (RSV) infection in a developing country. The work was carried out in three hospitals for primary cases and in the community for secondary cases in the western region of the Gambia, West Africa. RSV infection was diagnosed by immunofluorescence of nasopharyngeal aspirate samples in children younger than two years admitted to hospital with acute lower respiratory infection (ALRI). Routine records of all children with ALRI were analysed, and the incidence rates of ALRI, severe RSV-associated respiratory illness and hypoxaemic RSV infections were compared. A community-based study was undertaken to identify secondary cases and to obtain information about spread of the virus. 4799 children with ALRI who were younger than two years and lived in the study area were admitted to the study hospitals: 421 had severe RSV-associated respiratory illness; 55 of these were hypoxaemic. Between 1994 and 1996, the observed incidence rate for A...

The diagnosis of severe bacterial infection in young infants in developing countries is difficult because of the lack of sensitivity and specificity of the presenting symptoms and signs. Whether C-reactive protein (CRP) might help with the early detection of neonatal sepsis was investigated in a prospective study in The Gambia, Ethiopia and The Philippines. Infants < 3 months of age with symptoms or signs of possible sepsis were evaluated; CRP was measured and assessed for its ability to predict proven invasive bacterial infection. Of 966 children < 3 months of age, 54 had a positive blood culture, 13 a positive CSF culture, 15 a positive blood and CSF culture and 884 had negative cultures. Median (interquartile range) CRP values were 42 (9-173), 14 (6-36), 209 (135-286) and 8 (3-27) mg/L in the four groups, respectively. Taking a CRP cut-off of 10 mg/L, the sensitivity and specificity of an elevated CRP to predict a positive blood or CSF culture were 77% and 55%, respectively, and 55% and 82%, respectively, for a cut-off of 40 mg/L. CRP lacks the sensitivity and specificity to be used alone as a predictor of serious infections in young infants.

ABSTRACT We welcome the interest in our study1 assessing StopAdvisor, an internet-based intervention for smoking cessation, and thank Partha Haldar and Shashi Kant for their thoughtful critique. We would like to respond to the five issues that they have raised. Haldar and Kant's first point is that our study relied on an occupation-based measure of socioeconomic status, which was self-reported via an online questionnaire. The authors argue that online assessment is prone to error because people misreport personal information assessed over the internet. However, studies2, 3 have shown that misreporting does not necessarily occur during online assessment and, although socioeconomic status is difficult to assess without an interviewer being present, the measure used in our trial was explicitly derived and validated as a self-reported version of the National Statistics Socioeconomic Classification.4 Moreover, we were encouraged that the assessment in our trial showed internal validity by accurately representing characteristics associated with socioeconomic status. Finally, we note that although Haldar and Kant speculate about possible error resulting from our measure of socioeconomic status, they offer no explanation—and we can think of none—as to how error related to misreporting of socioeconomic status could have accounted for the pattern of results obtained. The second issue is that our measure of time spent on websites did not take into account times when users might have been doing other activities (including browsing other websites in separate windows). Such limitations with respect to the assessment of exposure to content are common in the assessment of digital interventions.5 As a proxy of general engagement with the websites, we noted convergent validity in that the results from the time measurement were similar to those for logins and page views. However, we agree with the authors that the finding that overall use of the website seemed greater by people with high compared with low socioeconomic status suggests that these crude indices of use did not mediate the effectiveness of StopAdvisor for smokers of low socioeconomic status. We are keen to investigate more nuanced measures of engagement in future research. The third issue relates to what is described as a so-called borderline result and the clinical significance of the effect size. Within classical inferential statistics, the result cannot be borderline—the results is either significant or not, and in this case it was significant. With respect to the effect size, we strongly disagree with the authors about its clinical significance. As described in our discussion, modest (and even small) effect sizes can be of great clinical significance because of the huge health gains associated with stopping smoking: effects as little as 1% on the proportion of people with 6 month sustained abstinence would result in at least 3 additional years of life for every 100 smokers aged 40 years given an intervention.6 We also discussed how the effect size for smokers with low socioeconomic status was similar to other modes of delivery for behavioural support. With respect to the fourth issue of possible duplicate users, an email address could only be used once on our website and was secured to the treatment allocation. Additionally, our eligibility criteria included the provision of valid telephone numbers and postal addresses for follow-up. We screened these personal details to ensure they were unique before participants were included in the trial, which resulted in the exclusion of 19 people (see figure 1 in the Article1). Finally, Haldar and Kant raise issues with our reporting of the analysis of the saliva samples. In fact, we did report the number of participants who did not provide verified saliva samples as a footnote to table 2 (207 [4·5%] of all 4613 participants for the primary outcome and 392 [8·5%] of 4613 for the secondary outcome, with similar numbers between intervention groups in each socioeconomic status subsample). Although we did not report descriptive statistics for the analysis of saliva samples, we did describe in detail the decision procedure for judgment of whether a saliva sample provided verification of the reported abstinence from smoking. Additionally, we do not see how our analysis of the saliva samples could have accounted for the pattern of results obtained.