Managua is the capital of Nicaragua and the largest city in the country; its estimated population is 1.4 million. The Health Center Sócrates Flores Vivas (HCSFV), a public primary care facility located in District II of Managua, served as the study site. HCSFV provides medical care for the ≈62,500 persons who reside in the surrounding catchment area.

We used information collected through the Nicaraguan Pediatric Dengue Cohort Study, a prospective cohort study established in August 2004 to study pediatric dengue infection. Recruitment of children 2–9 years of age was conducted by door-to-door visits in neighborhoods served by HCSFV. Children were ineligible to participate if their parents or guardians reported a history of any disease or treatment that might suppress the immune system, such as HIV/AIDS or treatment for cancer. To maintain the age structure of the cohort, during July and August of each year, additional participants 2 years of age were enrolled. Age range by the end of the 2-year study was 2–11 years. All participants were provided medical treatment and tests by study physicians at HCSFV free of charge, 24 hours per day, 365 days per year. A study ambulance was available at all times to transfer any child requiring hospitalization or emergency services not available at HCSFV. To determine the percentage of children who did not regularly attend HCSFV when ill and to encourage attendance, home visits were conducted at least 1 time per year. All children were contacted by study medical personnel each year during July or August; if the child could not be located after at least 3 attempts, the child was considered lost to follow-up.

Our analysis covered 2 years, from April 16, 2005, through April 15, 2007. As a part of the study enrollment process, study personnel used questionnaire forms that asked questions in a systematic way to gather demographic information and medical history. Subsequently, a household risk-factor questionnaire was administered at home visits and at HCSFV; children with completed household risk-factor questionnaires at the beginning or during the first few months of the time period analyzed in this study were selected as the convenience sample. The study was approved by the institutional review boards at the University of California at Berkeley and the Nicaraguan Ministry of Health. Informed consent was obtained from the parent or legal guardian of each participant, and participants >6 years of age completed assent procedures.

Parents or guardians agreed to bring participants to HCSFV at the first sign of fever and for all medical appointments. Medical information from all visits was recorded on forms for systematic data collection at the time of consultation. The Centers for Disease Control and Prevention (CDC) definition of influenza-like illness was used: fever >37.8°C at presentation with a cough and/or sore throat. ILI episodes were considered to be the same episode if they occurred within a week of each other, if they had the same date of fever onset, or if the physician denoted them as a continuing illness. High-probability ILI was defined as a fever >38.2°C at presentation with a cough (18). To eliminate the younger ages in which respiratory syncytial virus, a common seasonal respiratory virus, is often prevalent, a portion of the analysis was restricted to children >6 years of age (15,16).

Paired serum samples were collected for a subset of participants during each peak of ILI activity (11 paired samples from the 2005 peak, 15 from each of the 2 peaks in 2006, and 10 from nonpeak periods of each year); the first sample was collected at the initial visit, and the second was collected 2–4 weeks later. The hemagglutination inhibition test was performed by using the standardized reagents and protocols provided by the World Health Organization. Results were considered positive for influenza if a >4-fold rise in hemagglutination inhibiting antibody titer was noted between acute- and convalescent-phase samples. During December 2006–January 2007, a total of 51 nasal and throat swab samples were collected and analyzed by reverse transcriptase–PCR (RT-PCR). Swabs were collected from children who had fever or a history of fever and cough and/or sore throat and symptoms for <5 days. RNA was extracted from 140 μL of viral transport medium containing the swabs by using the QIAamp Viral RNA Isolation Kit (QIAGEN, Valencia, CA, USA) and amplified by using the Access RT-PCR System (Promega, Madison, WI, USA); primers were directed to influenza A and influenza B viruses according to the standard operating procedure of CDC’s Influenza Branch.

Total person-time was determined by the amount of time that a participant was enrolled in the cohort. For those lost to follow-up, person-time was determined by adding the known person-time between enrollment or the beginning of the study period and last contact with the study to half the amount of time between the last contact with the participant and the official loss date. Incidence was calculated as the number of ILI episodes divided by the person-time multiplied by 100. A Poisson distribution was used to calculate 95% confidence intervals (CIs) for the incidence rates. Weekly incidence was graphed and smoothed by using Lowess (19) with a 3-week moving average. Incidence rate ratios (IRRs) were used as the measure of relative risk. General estimating equations with a Poisson distribution were used to estimate IRRs. General estimating equations were chosen to account for the longitudinal nature of the study; most participants contributed 2 person-years of time. To check the assumption that the ILI events were uncorrelated, the models were also run with a negative binomial distribution; however, the results did not significantly differ from those of the models with a Poisson distribution, and therefore the Poisson distribution was used in all models. Robust standard errors with an exchangeable correlation structure were used to estimate the 95% CIs for IRRs. Model selection was performed by using a backward step-wise procedure. The χ² test was used to compare proportions. All analysis was conducted using STATA version 9.2 (STATA Corp, College Station, TX, USA).

A total of 4,276 children contributed 7,449 person-years of time to the study. Of these, 3,240 (75.8%) children contributed 2 full years of time, 555 (13%) 2-year-old children were enrolled during yearly maintenance enrollment, 118 (2.8%) children were withdrawn from the study, and 363 (8.5%) were lost to follow-up. Children were withdrawn from the study if their parents requested it, if they moved from the study area, or if they did not follow study procedures. Cohort characteristics are summarized in Table 1. Attendance at HCSFV was high; 83.3% of participants had >1 medical visit, and 94.1% of those with fever visited HCSFV by the fourth day after fever onset. At the time of home visits, only 1.9% of participants reported having sought medical care outside of the cohort.

A total of 2,596 episodes of ILI yielded an incidence rate of 34.8 episodes per 100 person-years (95% CI 33.5–36.2). Decreasing incidence was noted for each 1-year increase in age (Figure 1). Rates for boys and girls did not differ significantly (Table 2). Of the participants, 38% had had >1 episode of ILI; of those with ILI, the mean number of ILI episodes per child was 1.6 (range 1–8). Among children with ILI, the rate of hospital transfer for evaluation and possible admission was 14 per 1,000 person-years.

ILI episodes occurred with marked seasonality; they peaked during June–July in both years and again during November–December of the second year (Figure 2, panel A). The same pattern was present when a high probability of influenza definition was used (fever >38.2°C with cough) and when the analysis was restricted to children >6 years of age (Figure 2, panels B and C). Analysis of a subset of paired serum samples under each peak showed that all 3 peaks were likely due, at least in part, to influenza; 64% of the June 2005 peak samples were positive for influenza A virus (H3N2), 27% of the June 2006 peak samples were positive for influenza B virus, and 20% of the November–December 2006 peak samples were positive for influenza A virus (H1N1) and 7% for influenza B. All samples from nonpeak times were negative for influenza A and B viruses. RT-PCR of 51 samples from the November–December 2006 peak confirmed the hemagglutination inhibition results: 10 (20%) were positive for influenza A virus. Conversely, no influenza B virus-positive samples were identified among the samples collected in June 2005, and no influenza A virus (H3N2)–positive samples were detected in November–December 2006.

Demographic and medical risk factor information was available for all cohort participants. Young age was the strongest predictor of ILI; risk decreased with each increasing year of age (Table 3). Having received a prior diagnosis of asthma was a strong predictor of ILI; risk was 1.79 times higher for children with asthma (95% CI 1.56–2.06) than for other children, after adjusting for age and sex. No significant difference in the risk for ILI was noted between boys and girls.

Information concerning socioeconomic and household risk factors (hereafter referred to as household factors) was available for 2,615 (61%) participants (Table 1). The cohort and the subset of children sampled for household factors were similar with regard to sex distribution but differed slightly in age distribution and asthma status. Neighborhood geographic distribution of the entire cohort and of those with household risk factor data did not differ significantly (p = 0.98). Socioeconomic information on those surveyed is shown in Table 4. Household factors collected and included in the model were number of people living in the house, person density (determined by number of people living in the household divided by the number of rooms), mother’s literacy status, mother’s educational level, type of floor, and presence of a toilet or latrine. Electricity and running water in the house were not included in the model because >99% of the cohort had access to each (at least some of the time). The final model contained person density in the household, mother’s literacy status, type of floor, and the child’s age, asthma status, and sex (Table 5). A trend of increasing risk for ILI was noted with increasing person density in the household; homes with 3.0–4.9 persons per room had an RR of 1.07 (95% CI 0.96–1.20), and those with >5 persons per room had an RR of 1.18 (95% CI 1.04–1.34) compared with households with <3 persons per room. Having a literate mother was protective against ILI (RR 0.79; 95% CI 0.64–0.98). Although not retained in the final model, in the univariate analysis the mother’s educational level displayed a U-shaped pattern; risk decreased until completion of primary school and then increased with completion of secondary school or college.