Durability assessment results suggest a serviceable life of two, rather than three, years for the current long-lasting insecticidal (mosquito) net (LLIN) intervention in Benin

Background LLIN distribution, every three years, is a key intervention of Benin’s malaria control strategy. However, data from the field indicate that LLIN lifespan appears to vary based on both intrinsic (to the LLIN) and extrinsic factors. Methods We monitored two indicators of LLIN durability, survivorship and integrity, to validate the three-year-serviceable-life assumption. Interviews with net owners were used to identify factors associated with loss of integrity. Results Observed survivorship, after 18 months, was significantly less (p<0.0001) than predicted, based on the assumption that nets last three years. Instead, it was closer to predicted survivorship based on a two-year LLIN serviceable life assumption (p=0.03). Furthermore, the integrity of nearly one third of ‘surviving’ nets was so degraded that they were in need of replacement. Five factors: washing frequency, proximity to water for washing, location of kitchen, type of cooking fuel, and low net maintenance were associated with loss of fabric integrity. Conclusion A two-year serviceable life for the current LLIN intervention in Benin would be a more realistic program assumption.


Background
National distribution of long-lasting insecticidal (mosquito) nets (LLINs) is a proven malaria control intervention [1][2][3]. However, LLIN interventions have a limited serviceable life, and net replacement must be programmed in a timely way to maintain impact. Programs that wait too long risk operational failure; likewise, premature replacement is to be avoided for cost-effectiveness reasons. The current assumption regarding LLIN serviceable life is three to five years of use under field conditions [4,5]. While this has been incorporated into LLIN intervention planning, questions have arisen about whether such an approach holds true everywhere nets are in use [6]. Rather than a single LLIN serviceable life everywhere, data from the field indicate that LLIN loss appears to vary based on both intrinsic (to the LLIN) and extrinsic factors [7][8][9][10]. We monitored LLIN durability in Benin, during the current national LLIN distributionreplacement cycle, so that the timing of the future replacement campaigns can be informed by Benin-specific net loss data.

Methods
Beginning in July, 2011, approximately four million polyethylene-based LLINs (Olyset® net, Sumitomo Chemical Company) were distributed throughout Benin. Following distribution, a net tracking activity to monitor the durability of the LLINs [11,12] was implemented in four communities. Two of the communities, Kessounou and Allada, were located in Southern Benin (Figure 1), while the other two, Kandi and Malanville, were located approximately 750 km further North. Residents in Kessounou, located on the Oueme River, have ready access to water for washing nets. In contrast, residents of Allada, who want to wash a net, must carry water some five kilometers. Similar criteria (a short distance to water for washing LLINs versus a long distance) applied to the sites in the North. Malanville, located on the Niger River (water for washing nets easily accessible) and Kandi (residents must transport water for washing nets, as at Allada).
As described previously, at the time of the nationwide LLIN distribution in 2011, a sample of 2002 households, approximately 500 per site, was randomly selected [11]. Household selection at each site took into account all villages to ensure representative sampling [11]. Assessment teams identified a 2011-distribution LLIN in each selected household and, if the net was hanging and in use, enrolled it, and the household where it was located, in a LLIN durability monitoring assessment, based on World Health Organization guidelines [13]. Each selected LLIN was double tagged (a bar code, attached to the net, plus an indelible-ink symbol applied to the LLIN Figure 1 Reproduced copy (with copyright) of Benin Map LLIN tracking assessment sites: Kessounou, Allada, Kandi and Malanville [11]. fabric) to ensure correct identification during later visits. The GPS coordinates of the household and the name of the head of household, or an adult person acting on behalf of the head, were also recorded to facilitate follow up. This paper reported the 12 and 18 months assessments.

Monitoring survivorship/Attrition
Due to the way in which households were selected (one enrolled LLIN per household), LLIN survivorship at T 0 , was set at 100%. At 6-month intervals, each selected household received a follow-up visit. If the household was open at the time of the visit, the assessment team visually confirmed the continued presence of the coded LLIN. If the coded LLIN was not in the house, the assessment teams determined how the net went missing by interviewing the owner. Owners were asked to choose one of three reasons for why the net was no longer present in the household. These were: (i) the net was thrown away because it was physically damaged and thought to be of no value, (ii) the net was removed (e.g. given away, stolen, sold etc.), and (iii) the net was repurposed for an alternative use.

Monitoring integrity
LLIN fabric integrity was assessed by a visual examination, without removal of coded nets from selected households. Observed holes were assigned to one of four size categories: 1. a hole size of 0.5-2.0 cm or '<a thumb-sized opening' 2. a hole size of 2.0-10.0 cm or '>a thumb but<a fist' 3. a 'hole size of 10-25 cm or '>a fist but<a head' 4. a hole size of >25 cm or '>a head' The most-likely cause of the damage, a rip in the fabric, a rip in the seam, burned burn-related hole or the result of rodent damage was also recorded.

Interview questionnaire
A questionnaire, developed by WHO [13], to identify factors associated with survivorship, was adapted for use in the assessment. Questions were programmed (ODK Collect 1.2.2 software) into Samsung Galaxy Tablets to record responses.

Data analysis Survivorship
The equation for quantifying overall survivorship, also referred to as attrition, was: Total coded LLINs still present in the households selected Total coded LLINs at enrollment T 0 ð Þ Â 100 If a household was closed, during an assessment visit, it was treated according to the non-parametric survival method of Kaplan-Meier [14]. Survivorship, plotted against time (T 6 , T 12, T 18), was compared with NetCALC net loss model curves based on 2-year and 3-year LLIN serviceable life assumptions (http://www.networksmalaria.org). Equations for calculation of LLIN survivorship/attrition associated with three different reasons for why an assessment net had gone missing were: Attrition rate-1 (reason: physical damage): Total number of coded LLIN reported as thrown out due to wear and tear in surveyed households Total coded LLINs at enrollment T 0 ð Þ Â 100 Attrition rate-2 (reason: removal): Total number of coded LLIN reported as given away; stolen; sold or used in another location Total coded LLINs at enrollment T 0 ð Þ Â 100 Attrition rate-3 (reason: re-purposed): Total number of coded LLIN reported as being used for another purpose in surveyed households Total coded LLINs at enrollment T 0 ð Þ Â 100 Two communities were reported to show significantly different survivorship/attrition if the 95% confidence limits did not overlap.
Integrity was quantified based on two measurements: 1) The proportion of LLINs with any hole.
Total number of coded LLINs with at least one hole of size 1-4 Total number of coded LLINs found and assessed in surveyed households Â 100 2) The proportionate holes index (pHI) for each net [13] 1 × number of size − 1 holes + 23 × no. of size − 2 holes + 196 × no. of size − 3 holes + 576 × no. of size − 4 holes. The figures, 23, 146, and 576 refer to the estimated mean hole area for the different sized holes. Descriptive statistics were used to compare pHI values at each assessment site (mean, median, interquartile range). Based on the pHI score, LLINs were assigned to one of three condition catagories (Roll Back Malaria: Measurement of Net Durability in the Field: Current Recommended Methodology, presented in Lyon, February 2012).
pHI ≤64good pHI ≤768serviceable pHI >768replace Factors associated with loss of integrity were identified by multivariate regression analysis of nets in the 'replace' category and frequency of responses (by owners of the nets). Modalities with very low numbers observed were aggregated with those that have high numbers for the multivariate analysis.

Study clearance
This prospective study was planned with and approved by the Ministry of Health. Community leaders were informed before the study and all gave verbal consent before initiation. Written consent was then obtained on the day of the study from all participating households.

Net survivorship/attrition
There were 2002 nets enrolled (T 0 ) in the assessment. During the T 6 T 12 and T 18 follow up visits, 1672, 1225, and 973 LLINs, respectively, were found and evaluated (Table 1). There was a significant difference in survivorship associated with community location, but not associated with distance to water for washing nets. After  for the 'near' (to water) ones. In summary, survivorship (all communities) was 93% after 6 months, 72% after 12 months and 57% after 18 months; was variable between southern and northern localities; but did not appear to change based on distance to water for washing (LLINs loss in South was 01.62 [CI 95%: 01.41-01.86] times that in the North, but the same at sites near water versus those located farther away. Observed survivorship is compared with NetCALC loss predictions (NetCALC loss curves) in Figure 2. Net-CALC predicted survivorship at 18 months is 84% for the three year model, and 68% for the 2-year model, whereas observed survivorship in all communities at 18 months, 57%, is significantly less (P<0.0001) than predicted based on the three-year assumption and also less than, but closer to, that predicted by the NetCALC 2year loss rate model (p=0.036).

Reasons for net loss
There were 313 and 289 interviews, at households where the net was missing, administered during the 12 and 18

LLIN fabric integrity
Of the nets remaining in the households to which they were distributed, the percentage with any hole, ranged from 62% at Allada to 87% at Malanville and Kessounou at T12 (Table 3). At T18, the percentage of LLINs with any hole, ranged from 72% at Allada to 93% at Malanville. Consistent differences in measures of integrity were observed when results for the two communities that were near the source of water for washing nets were compared with the two communities that were farther away from the source of water for washing (Table 3).
By the end of one year (T12), the mean proportionate hole indices in the communities located near water for washing (area of greater washing frequency) were 691 and 799, versus 398 to 510 in the communities located farther from washing (lower LLIN washing frequency). Additionally, median pHI values for the communities nearer to the 'wash water' source, 290 and 243, were three to six times greater than that observed in communities that were farther away from water for washing nets, 86 and 46. Applying the net condition categories, 'good' , 'serviceable' and 'needs replacement' , 33 and 30 percent of the nets in communities near water for washing were in need of replacement versus 16 and 22 percent in the communities that were farther away from water. Furthermore, the number of LLINs in the 'good condition' category was significantly higher (p<0.05) in location with less access to water. At the 18-month assessment visit, the mean proportionate hole indices in the communities located near water for washing (area of greater washing frequency) were 888 and 1479, versus 623 to 447 in the communities located farther from washing (lower LLIN washing frequency). Median pHI values for the communities nearer to a 'wash water' source, 706 and 312, were also greater than that observed in communities that were farther away from water for washing nets, (105 and 96). There were 48 and 36 percent of the nets in communities near water for washing in need of replacement, versus 24 and 18 percent in the communities that were farther away from water (Table 3). LLINs in the 'good condition' category were also significantly higher (p<0.05) in location with less access to water for washing nets. When LLINs with any holes were categorized by the nature of hole as representing either a rip in the fabric, a rip in the seam, a burn hole or the result of chewing by rodents, the 'rip in the fabric' category accounted in average for 85% of the damage, while burn holes accounted for 10%, open seams for 3% and rodent damage for 2% at T12 assessment (Table 4). At T18 assessment, the 'rip in the fabric' category accounted in average for 84% of the damage, while burn holes accounted for 11%, open seams for 3% and rodent damage for 2% (Table 4). Factors that showed a significant relationship with loss of integrity as measured by nets with any hole included: washing frequency, LLINs maintenance (low), location of the kitchen (inside the house), type of cooking fuel and the distance to water for washing (Table 6). High washing frequency increased the risk of physical damage to the LLINs (p<0.0001).

Discussion
LLIN loss, measured by survivorship/attrition, occurred more rapidly than predicted by the 'three-year serviceable life' assumption, currently used in Benin to program distribution/replacement of LLINs. Of interest, however, was the observation that survivorship rates did not appear to be affected by LLIN washing frequency. In contrast, our results showed a marked effect of washing frequency on the integrity of the LLINs. Therefore, LLIN loss of physical integrity/deterioration, a factor that affects the ability of LLINs to prevent mosquito-human contact, may not be taken into account if only survivorship is monitored. In this evaluation, loss of integrity, measured in LLINs that remained in place, was so extensive that it may well have compromised the value of the LLINs still remaining in the households as a malaria prevention measures. Recent work [15] in which damaged LLINs are shown to increase man-vector contact from none to an average of five bites/man/night demonstrates the impact that a relatively small loss of integrity,  T12  T18  T12  T18  T12  T18  T12  T18   Tagged LLINs found  253  227  338  279  286  184  348   Net removal for all reasons was observed in 6% of the study houses at T6, in 18% at T12 and in 26% at T18. Some, perhaps most, of these nets were moved by design of the household (e.g. given away to other houses in the community) and therefore, it is possible that the nets continued to contribute to community protection. Nonetheless, the percentage of nets thrown away (25%) and nets re-purposed were surprisingly high in some communities, eg. Kessounou.
The proportionate hole index (pHI) [13] provides a standardized approach to describing changes in LLIN fabric integrity. Applying pHI thresholds [16]: 'like new' , 'needs repair' and 'needs replacement' , to our results, we observed that after 12 months, 16-22% of the LLIN need to be replaced at locations with less access to water (Allada and Kandi) versus 30-33% at locations with ready access to water for washing (Malanville and Kessounou). Estimates of net loss associated with integrity (computing loss as proportion of nets in pHI category 'needs replacement' were between 18% and 48% after 18months). If loss associated with fabric integrity declines at these rates, more than one half of the LLINs distributed in 2011, and still present in the household to which they were given, would need replacement prior to two transmission seasons post-distribution. Thus, based on fabric integrity alone, the LLINs would not be expected to provide adequate protection after two transmission seasons.
Cooking fuel, location of the kitchen as well as low LLINs maintenance and washing frequency were correlated with the loss of fabric integrity. Social practices regarding net care and repair, vary from community to community, and most likely influence condition of the LLINs in each region [17,18]. It important, therefore, to reinforce awareness about the best practices related to net care and repair.
A recent study in Kenya [19] reported that people washed their LLINs more frequently than recommended and associated this practice with poor physical quality of nets. The study also noted that light colored nets were more likely to be washed than were dark colored nets. The polyethylene-based LLINs in this assessment were light blue. The majority of the nets (54-74%) were observed to be dirty. Thus, it may be that these factors also contributed to the high observed frequency of washing/rapid decline in fabric integrity. They could significantly affect the bio-efficacy of the nets that decreased from 9-58% after 6 months [11]. However, bio-efficacy results were not included in the present assessment-and represent a potential limitation of the study.
The findings of this study have important implications for the LLIN-malaria control strategy in Benin. They suggest that: the current polyethylene-based LLIN, distributed during 2011, has an effective life of two rather than three years; that behaviour change communication (BCC) strategies that support LLINs repair and that discourage subsequent removal of nets after distribution should be strengthened in an attempt to sustain higher coverage levels for a longer period of time; and that channels for routine replacement of LLINs, between national campaigns should be strengthened to replace nets that no longer meet minimum fabric integrity standards; and that other WHOPES-approved LLIN products should be evaluated to determine which is most costeffective. Finally, it may be that local communities can adopt and use WHO guidance on assessment of integrity (counting holes) and survival to assess LLIN loss, thereby verifying ongoing impact under local conditions.