This investigation was unusual in that net evaluation began very shortly after distribution. Since LLINs are expected to last for multiple years, most prior studies have been of nets a year or more after distribution. However, for programme planners preparing for replacement campaigns, it would be useful to have meaningful data about nets in the field and their rate of deterioration as early as possible. Results from this study showed that in as little as three to six months, a quantifiable picture began to emerge regarding the physical deterioration of nets in the field. Future LLIN monitoring efforts can therefore start less than a year after distribution in order to give planners a head start on developing timetables for replacement.
Although some nets were still in good physical condition even after a year-and-a-half in this study, a substantial proportion showed significant deterioration: 68% had holes and 28% were classed as ‘mediocre’ or ‘poor’ by hole index at 14 to 20 months of use. The distribution of holes per net was highly skewed, with a few nets having many holes and most nets with few holes. For this reason, statistical descriptions of overall condition of a group of nets are better described by the median and not the mean number of holes, the latter being disproportionately impacted by the existence of a few heavily damaged nets. However, the size of holes must be taken into account as well: one large hole may be as problematic (or even more so) than a large number of small holes, since large holes can let more mosquitoes inside of the net without contacting the insecticidal fibres.
The two schemes proposed [9, 10] to estimate a standardized hole index differ in both the number of size categories and the relative factors applied to each. In the first scheme, WHO  recommended four size categories of 0.5 to 2 cm, 2 to 10 cm, 10 to 25 cm and >25 cm, with midpoint hole diameter of 1.25, 6, 17.5 and 30 cm respectively. The relative multiplication factors based on corresponding hole areas for these groups are 1, 23, 196 and 578. Secondly, Kilian et al. and Batisso et al. used three size categories <2 cm, 2 to 10 cm and >10 cm (finger, fist and head). They estimated average hole areas to be 4, 36 and 225 sq cm respectively, giving multiplication factors of 1, 9 and 56 for small, medium and large holes to arrive at the proportionate hole index. The current study commenced before either of these schemes was published, but elected to use the method of Kilian et al.  and Batisso et al.  for hole index calculation as it is based on three size categories similar to those used here, and to enhance comparability between studies in Ethiopia.
A limitation of this study is that net attrition (complete loss of nets through disposal, diversion to other use, sale, or donation to others outside the household) was not measured, so net deterioration and loss may have been underestimated. Attrition has been estimated to be as high as 32% over three years (Batisso et al. ). Other limitations include the convenience sampling method for the sites and households, and the fact that the sites were at a range of altitudes, so net use (and as a result, net wear) may differ significantly depending on mosquito populations and perceived risk of malaria.
Repairs to nets that had developed holes were rarely observed in this study. Low repair rates were also reported by Smith et al. , Kilian et al.  and Shirayama et al.  so this appears to be a widespread issue. However, Bhatt et al.  found a total of 750 repairs, or average 1 repair per LLIN, in the forms of stitches (63.9%), knots (35.8%) and patches (0.3%) in their evaluation of Interceptor® LLINs (n = 932) in central India. It is likely that longevity of nets can be significantly improved simply by making repairs to them and perhaps by extension, encouraging behaviour that would prevent the nets from developing holes to start with.
Although the cause of the holes in the net studied here is not known, household observations suggested that fire embers together with damage from animals and sharp edges on beds or mats were major contributors. This will be explored further in risk factor analysis, using hole index and insecticide concentration as outcomes, for the 2008 and 2009 collections in a subsequent paper. Holes were observed on all sides of the net including the top panel, indicating that all parts of the nets are susceptible to damage, but net density was greater in the lower sides of the net. Of note, the relative proportion of small, medium and large holes did not change over time. Thus the hole size (and hole index) distribution at a future date could possibly be predicted using data from nets used for only a short period. Furthermore, if the proportion of small, medium and large holes is different for different net brands, it could become a measure of net quality (quality increasing as large:small hole ratio decreases). Knowing the ratio of large to small holes could have a profound impact on program planning and procurement decisions, and it would be desirable for future similar durability studies to report on this ratio.
Comparison of the physical deterioration rates between studies is difficult because of differing methodologies and time periods of assessment. Kilian et al.  observed that for PermaNets in Uganda, >70% had holes in them after one year (i e, <30% were undamaged). In this study, net attrition was not measured, as in the current study. Although nets were not collected after exactly one year, the findings in the current study were fairly consistent: it was observed that 54.5% of nets were damaged after three to six months and 85.5% after 14 to 20 months. In Tanzania, Maxwell et al.  defined “intact” nets as having <20 holes <2 cm in diameter, <5 holes 2–5 cm in diameter, and <2 holes >5 cm in diameter. Maxwell et al. found that 84.4% of Olyset nets in highland villages were ‘intact’ after four to five years and 38.4% of the nets in lowland villages were ‘intact’ after six to seven years . In the current study, the time period is shorter and the hole size comparison is not strictly similar. However, it was found that 65.5% of nets had <20 holes (any size) at 14 to 20 months and 46.8% had <20 holes at 26 to 32 months, suggesting that these PermaNet®2.0 were deteriorating more quickly than the Olyset nets observed in Maxwell’s study.
In Chad, Allan et al.  used a proportionate hole index to assess LLIN durability in Interceptor, Olyset, and PermaNet after one year of household use, and assessed damage as holes < 2 cm, >2-5 cm and >5 cm. A total of 876 LLINS were assessed and 25% had a hole index greater than 300; these nets were classified in their study as “unserviceable and irreparable”. The largest percentage of nets (i.e. 44.5%), classified as “in partial service” had proportionate hole indices in the middle ranges of 25–174 and 175-299 . The most relevant comparison for our study is to that of Batisso et al.  in Ethiopia, who observed that after three years, 70% of nets were still in ‘good’ or ‘fair’ condition as defined by a hole index of <175, whereas in this study the percentage overall in these two categories was lower at a slightly earlier time period: 50.5% at 26 to 32 months. As noted by Batisso et al., work still needs to be done to determine when nets should be classified as ‘unserviceable’, as well as comparisons of net life in different circumstances within the same country and between countries.