Estimation of community-level influenza-associated illness in a low resource rural setting in India

Abstract

Objective: To estimate rates of community-level influenza-like-illness (ILI) and influenza-associated ILI in rural north India.

Methods: During 2011, we conducted household-based healthcare utilization surveys (HUS) for any acute medical illness (AMI) in preceding 14days among residents of 28villages of Ballabgarh, in north India. Concurrently, we conducted clinic-based surveillance (CBS) in the area for AMI episodes with illness onset ≤3days and collected nasal and throat swabs for influenza virus testing using real-time polymerase chain reaction. Retrospectively, we applied ILI case definition (measured/reported fever and cough) to HUS and CBS data. We attributed 14days of risk-time per person surveyed in HUS and estimated community ILI rate by dividing the number of ILI cases in HUS by total risk-time. We used CBS data on influenza positivity and applied it to HUS-based community ILI rates by age, month, and clinic type, to estimate the community influenza-associated ILI rates.

Findings: The HUS of 69,369 residents during the year generated risk-time of 3945 person-years (p-y) and identified 150 (5%, 95%CI: 4-6) ILI episodes (38 ILI episodes/1,000 p-y; 95% CI 32-44). Among 1,372 ILI cases enrolled from clinics, 126 (9%; 95% CI 8-11) had laboratory-confirmed influenza (A (H3N2) = 72; B = 54). After adjusting for age, month, and clinic type, overall influenza-associated ILI rate was 4.8/1,000 p-y; rates were highest among children <5 years (13; 95% CI: 4-29) and persons≥60 years (11; 95%CI: 2-30).

Conclusion: We present a novel way to use HUS and CBS data to generate estimates of community burden of influenza. Although the confidence intervals overlapped considerably, higher point estimates for burden among young children and older adults shows the utility for exploring the value of influenza vaccination among target groups.

Fig 1. Schematic diagram depicting estimation of person risk-time.

A. shows estimation of risk-time and apportioning by month. For each surveyed household, all the members contributed risk-time of 14 days preceding the date of interview. Risk time for any month was estimated by summation of person-days of risk-time contributed for that month by members of each household surveyed. B. shows timeline of surveys and month-wise risk time in person-years. The solid columns indicate number of households visited and striped columns indicate number of persons surveyed. The shaded area depicts the risk-time in person-years contributed by persons surveyed and apportioned by months.

Fig 2. Flow-chart depicting study methods for estimation of rates of influenza-associated ILI* in 28 villages in the Ballabgarh block (Haryana, India)–January 1 to December 31, 2011.

(*ILI: influenza like illness defined as cough with history/measured fever. **AMI: acute medical illness defined any illness irrespective of symptoms excluding injury and those related to obstetric or surgical problems. ***p-y: person-years).

Fig 3. Rates and 95% confidence intervals per 1,000 person-years of influenza-like illness (ILI) and influenza-associated ILI by age, Ballabgarh, India, 2011.

(Vertical axis denotes rates per 1000 person-years and horizontal axis denotes age-groups. The ILI rates are shown in blue striped columns while influenza-associated ILI is shown in purple solid columns. Whiskers indicate 95% confidence intervals).