Assessing bed net damage: comparisons of three measurement methods for estimating the size, shape, and distribution of holes on bed nets

Abstract

Background: Measuring the physical condition of long-lasting insecticidal nets (LLINs) under field conditions is of great importance for malaria control programmes to guide decisions on how frequently to replace LLINs. Current guidelines by the World Health Organization Pesticide Evaluation Scheme (WHOPES) propose a proportionate hole index (pHI) for assessing LLIN condition by counting the number of holes the size of a thumb, fist, head, and larger than a head. However, this method does not account for irregular hole shapes or exact hole sizes which could result in inaccurate decisions about when to replace LLINs.

Methods: LLINs were collected during a 2013 health facility-based malaria case control study in Machinga District, Malawi. To evaluate the accuracy of the pHI, the physical condition of 277 LLINs was estimated by the WHOPES method and then compared with two more thorough measurement methods: image analysis of digital photographs of each LLIN side; and for 10 nets, ruler measurements of the length, width, and location of each hole. Total hole counts and areas per net were estimated by each method, and detailed results of hole shapes and composite pictures of hole locations were generated using image analysis.

Results: The WHOPES method and image analysis resulted in similar estimates of total hole counts, each with a median of 10 (inter-quartile range (IQR) 4-24 and 4-23, respectively; p = 0.004); however, estimated hole areas were significantly larger using the WHOPES method (median 162 cm², IQR 28-793) than image analysis (median 13 cm², IQR 3-101; p < 0.0001). The WHOPES method classified fewer LLINs in 'good condition' compared to image analysis (42% vs 74%). The ruler method detected significantly more holes than image analysis did (p = 0.002) in 10 LLINs; however, total hole area was not significantly different (p = 0.16). Most holes were not circular but roughly 2-5 times longer in one direction. The lower quarter of LLIN sides was found to have the most holes.

Conclusions: The WHOPES method overestimated total hole area, likely because holes are elongated rather than circular, suggesting further adjustments to the pHI formula may be warranted when considering LLIN replacement strategies.

Keywords: Bed net; Hole index; Image analysis; Malaria; Physical durability; Vector control.

Fig. 1

Photograph of a study net hung on a frame. Image of the bottom half of a short side of a study LLIN

Fig. 2

Total hole counts (a) and total hole area (b) as measured using WHOPES assessment and image analysis methods. Scatterplot displaying the a number of holes or b total area of hole damage in each LLIN (represented by a dot) as measured by the WHOPES methodology (y-axis) and image analysis (x-axis). The dashed line represents a fitted line assuming a simple linear regression model. The solid diagonal line represents y = x. Kernel density plots on the margins show a smoothed distribution of counts (or area) by each method

Fig. 3

Aspect ratio of holes measured using image analysis. Jitter plot with box and whisker plot overlay displaying the distribution of the aspect ratio of holes as measured by image analysis (shown separately for the different WHOPES hole size categories, d = diameter). The jitter plots show the distribution of aspect ratio (each hole is represented by a dot). The box in the box and whisker plot show the 75th percentile, median, and 25th percentile, respectively; whereas the whiskers identify the extremes, including the minimum and maximum. Aspect ratio is the ratio of the major axis to the minor axis

Fig. 4

Composite visualization of hole damage by net side. Composite visualization of hole damage for the a long, b short, and c roof sides of study LLINs. Holes (represented as ellipses) are plotted based on measurements from image analysis (centroid, area and aspect ratio) for each net side, and sides of the same type were overlaid to form a composite image

Fig. 5

Hotspot map of hole damage by net side. Hot spot analysis of the composite images of hole damage using the Getis-Ord Gi* statistic for the a long, b short, and c roof sides of study LLINs. This statistic identifies areas (grids) on the LLIN side with statistically significant higher (red) or lower (blue) than expected counts (incidence) of holes on the LLIN side compared to a random process