Long-lasting insecticide nets (LLINs) are now entrenched as a major anti-malarial intervention [1]. In most malaria-endemic countries > 50% of households own at least one insecticide-treated net (ITNs) [1]. Despite this extensive coverage and use of ITNs over the last decade, their effective lifespan, and especially their physical integrity, is not well-understood under true operational and varied epidemiological settings [2–6]. While there is standard method of quantifying the number of holes in a bed net, lack of a standardized method to define a functional bed net in the field is an important limitation of studies evaluating the effectiveness of bed nets [2].

Bed nets reduce human-vector contact by providing a physical barrier between the human sleeping under the bed net and the malaria vector mosquito. This protection is enhanced when the bed net is treated with an insecticide that deters, repels, or kills vectors that attempt to bite the sleeper. A bed net that has its mesh (the number of holes per square inch) intact will rarely allow mosquitoes to reach the person sleeping under the bed net [7]. However, with use bed nets accrue holes that are big enough to allow mosquitoes to pass through. Thus, a functional or useful net is one that is physically intact and has insecticidal protection [8]. In addition to evaluating the physical presence (retention) of nets, it is paramount for malaria control programmes to also investigate the ‘usefulness’ of surviving nets (nets under routine use) [8]. Recent publication of field guidelines on collection of hole data is commendable [9]. However, these guidelines provide only a standard approach of quantifying the number of holes on a net, but a measure of categorizing the holes in to ‘effective net’ (likely to fulfill their protective function) and ‘ineffective net’ (offering diminished or no protection from mosquito bites) is lacking [2].

Not surprisingly, most controlled experiments on the importance of holes on bed nets have concluded that whether treated or untreated, protection offered by bed nets diminishes with increasing number of holes [10–12]. In Tanzania, Olyset nets (LLINs) after seven years of use provided personal protection only when they were in good physical condition [13]. Reduced effectiveness of bed nets in poor physical condition has also been reported in different epidemiological settings. Untreated worn nets (≥5 holes of about 2 cm in diameter, or ≥5 holes of unknown in diameter) offered significantly reduced protection against malaria in rural Gambia and in Kilifi, Kenya [14, 15]. In a recent study, untreated nets with holes were not protective, and untreated nets with no holes offered more protection against malaria infections than holed LLINs/ITNs [16]. Diminished protection from bed nets is further exacerbated if the vectors are resistant to the pyrethroids used in bed net [10, 11]. Moreover, an increasing number of studies suggest that the frequency of bed net use decreases with increasing net age and increasing physical damage of nets [17–19].

Published studies on physical integrity of bed nets in Uganda suggest that considerable physical damage (45%-78% of damaged nets) can occur even within a year of bed net use in operational conditions [3, 6]. Recently, Githinji and colleagues [20] reported poor physical condition in 40% of nets a year after distribution. Physical damage to nets is linearly correlated with reported bed net age and most bed nets are no longer in use beyond two years of use [2, 5, 6, 21]. Under ‘real life’ conditions, LLINs are expected to have two to three years of useful life [22]. Nevertheless, various researchers have variably estimated the useful life of bed nets to be up to four years [4, 5] As a result of the insecticidal effect on mosquitoes, insecticide treated nets have slightly longer useful life than untreated nets [4, 5, 23]. Regardless of insecticide treatment status, the rate at which holes appear on the net may be predisposed by net fabric [24] and fibre weight or denier [5]. Other factors that influence physical deterioration are the house environment (house wall material, bed type and construction), social economic status (SES) and the bed net maintenance behaviour (general handling, washing and repair) [6, 9]. Animals (goats, sheep, cattle, rodents, cats), fire from various sources (oil lamps, sparks from cooking, candles) and snagging on the bed frame are some of the known causes of holes on bed nets [5, 25].

Following the free mass distribution of bed nets in 2006, bed net ownership in Kwale County has been sustained at very high levels [26–28]. The distribution of free nets at maternal and child clinics to pregnant women and children under five years of age has been the primary approach through which most nets continuously get into the community. Secondary approaches include social marketing, where partial subsidy is offered, and ad hoc distribution by NGOs. Bed nets are also available through retail outlets at full cost for those who can afford them. Although the ownership situation continues to improve, the physical condition and maintenance behaviour of the nets in use is rarely evaluated and may undermine efforts to scale up ITNs. This paper describes the physical condition and identifies some of the predictors for poor physical condition of bed nets owned by individuals from communities in Kwale County, coastal Kenya. Additionally, net maintenance behaviour and presence and reported age of other nets that were not in routine use were documented.

The survey included interviews with household heads and inspection of bed nets in 1176 households with 5526 individuals and 2,786 nets. Of the 2,786 nets, 1,849 (66%) were in routine use while 937 (34%) were not. Eighty percent (942/1176) of the households owned at least one net that was routinely used. Nets under routine use were further categorized as effective or ineffective nets. Overall, despite 82% of the total people who slept in the households surveyed having had any net and over 96% of nets found hanging, only 63% of the people slept under a net the night prior to the survey. The 937 nets not under routine use were identified in 49% (576/1176) of the households surveyed. These were further categorized as misused nets, physically damaged nets, extra nets or acceptability nets.

Relationship between bed net age; person-type and physical condition of nets in routine use

Respondents reported knowing the age of 95% (1,753/1,849) of all bed nets. The reported mean age of the nets was 2.4 years (SE: 1.5, Range 0–10, Median: 2). Proportions of bed nets age used by different person-types are shown in Table 4. The number of bed nets used decreased with increasing bed net age for bed net 0–3 year old. The impact of the 2006 mass distribution of bed nets was evident; there were slightly more 4 year old bed nets than 3 year old ones. The rate of physical deterioration increased linearly with increasing bed net age (Figure 2).

Demographic category	≤1y (%)	2y (%)	3y (%)	4y (%)	>4y (%)	Mean age (SE)
Pregnant women	48.5	9.1	15.1	18.2	9.1	2.2 (1.6)
Infant (0-1yrs)	67.1	11.8	9.2	7.9	4.0	1.6 (1.3)
Young children (2-5yrs)	39.8	20.8	14.4	17.6	7.4	2.3 (1.4)
Older children (6-18yrs)	30.6	21.6	11.5	23.9	12.4	2.6 (1.5)
Others (above 18yrs)	39.0	21.0	12.9	17.0	10.1	2.0 (1.5)
Total	36.3	20.6	12.7	19.2	11.2	2.4 (1.5)

The age of the bed nets varied significantly with the person-type (Kruskal-Wallis χ² = 75.76, p < 0.0001). Post-hoc analysis showed that the age of bed nets owned by young children, adults and pregnant women was comparable (Table 4). Infants owned the newest bed nets compared to all other person-types, while adults owned newer bed nets than older children. Consequently, the physical condition of bed nets varied significantly between different demographic categories (Kruskal-Wallis χ² = 65.73, p < 0.0001). Bed nets used by infants and pregnant women were in the best physical condition (Figure 3F). Young children used bed nets in better condition compared to older children and adults, but the difference with adults was just above the 0.05 significant level (χ² = 3.82, p < 0.06). Bed nets used by older children were in the worst physical condition, while adult bed nets were only better relative to those owned by older children (Figure 3F). Accordingly, not only did the proportion of people protected by ‘effective nets’ vary with bed net age but also with demographic category (Figure 3A-E). Among pregnant women, infants, young children, older children and adults, 62%, 71%, 50%, 41% and 55% owned ‘effective nets’ respectively.

Among those who owned ≤1year old bed nets, the proportion of people protected by ‘effective nets’ was 92%, 88%, 86%, 80%, and 86% for pregnant women, infants, young children, older children and adults respectively (Figure 3A-E). This trend held across the other bed net ages (Figure 3A-E). For bed nets ≥2 years old, the proportion of people protected by ‘effective nets’ was 41%, 38%, 35%, 26% and 40% for pregnant women, infants, young children, older children and adults respectively

Environmental setting for bed nets in routine use

The age of bed net used in each environmental setting differed significantly (Kruskal-Wallis = 54.93, p < 0.0001).The age of bed nets in years in the coastal plain (mean: 2.81) and estuarine (mean: 2.88) environments were similar (p > 0.5) but significantly different from the coastal slope (mean: 2.33) and semi-arid inland (mean: 2.10) environments (p < 0.0001). Bed net age was not statistically different between bed nets found in the coastal slope and semi-arid inland environments (p > 0.05). The physical condition of the bed nets in the different environmental settings did not correspond to their ages.

As would be expected, bed nets in the estuarine environment were in worse physical condition; followed by semi-arid inland, and coastal plain while the coastal slope had the least damaged bed nets (Figures 1 and 4). In the coastal slope, coastal plain and semi-arid inland the frequency of bed net use (5–7 nights/week) was 77%, 90% and 95% respectively, suggesting sustained use resulting in faster physical deterioration. However, this observation did not hold for the estuarine environment where bed net use frequency was 93%, indicating that other factors contributed to bed net poor physical condition.

Predictors of physical condition of bed nets in routine use

Results of the best GAM logistic model are presented in Table 5. The first eight models with a ωi >0.01 are shown in Table 6. The best model (model number one; with the lowest AIC value and highest ωi) included as predictors the age of the bed net, the washing and use frequency, the person type, the net color, and the environmental setting. Physical deterioration of nets occurred with increasing bed net washing frequency (OR = 0.89, 95% CI: 0.81-0.93). Physical deterioration of bed nets was positively associated with increasing frequency of use but this association was not significant for bed net use of 3–4 nights or 5–7 nights per week in comparison to 1–2 nights. Relative to infants, young children and older children were found to use the ineffective bed nets significantly more often, but the physical integrity of nets owned by pregnant women was similar to those owned by infants (Table 5). People in the estuarine environment owned nets in the worst physical condition while those in the coastal slope owned bed nets in fairly good physical condition. As a result of the distribution strategies (most nets were of same brand and sourced through the maternal and child clinics distribution), most nets were similar in terms of net fabric, colour, size, brand and shape; with univariate analysis showing similar trend in their association with pHI. Thus, only net colour was included in logistic regression. Different from other predictors, the age of the bed net was included in the GAM model as a non-linear predictor. It thus possible to calculate the odds of an ‘effective net’ after every year of net use. The odds ratio of the age remained positive from 0 to 1½ years and reached the peak of negative value in the half of the second year of use (Figure 5). Other than a slight increase during the 4^th year of bed net use due to free mass distribution that was done four years prior to this study, the odds ratio of the age remained almost constant from 3^rd year onwards.

Predictors	Odds ratio (95% CI)
Bed net washing frequency	0.89 (0.81-0.93)*
Person-type (Ref: Infants)
Young children (2-5yrs)	0.54 (0.34-0.87)*
Older children (6-18yrs)	0.42 (0.30-0.69)**
Adults (above 18yrs)	0.64 (0.43-1.03)
Pregnant women	1.05 (0.44-2.36)
Bed net use frequency (Ref: 1–2 nights/week)
3-4 nights/week	0.57 (0.28-1.08)
5-7 nights/week	1.15 (0.71-1.58)
Net colour (Ref: White)
Green	0.81 (0.55-1.20)
Blue	1.60 (1.23-1.89)**
Environmental setting (Ref: Estuarine)
Coastal plain	2.31 (1.84-3.01)
Coastal slope	3.45 (2.40-4.16)**
Semi-arid inland	2.10 (1.39-2.57)**

Significant predictors of an ‘effective net’ are in bold (* = p < 0.05, ** = p < 0.01)

The odds ratio represents the probability of owning an ‘effective bed net’.

Model number	Age	Person type	Bed net colour	SES	Ecological setting	Frequency of use	N people	Washing frequency	df	logLik	AICc	delta	ωi
1	+	+	+	-	+	+	-	+	15	−1429.68	2889.55	0	0.52
2	+	+	+	-	+	+	-	+	14	−1431.94	2892.05	2.50	0.15
3	+	+	+	-	+	+	+	-	14	−1431.95	2892.08	2.52	0.14
4	+	+	+	-	+	+	-	-	13	−1433.95	2894.05	4.49	0.05
5	+	+	+	+	+	+	+	+	19	−1428.19	2894.70	5.15	0.04
6	+	+	+	-	+	-	+	+	13	−1434.55	2895.24	5.68	0.03
7	+	+	+	+	+	+	-	+	18	−1430.52	2897.32	7.77	0.01
8	+	+	+	+	+	+	+	-	18	−1430.55	2897.38	7.83	0.01

In this study, the physical condition of bed nets of various ages as reported by net owners in a malaria endemic region in coastal Kenya was examined. At least one hole was observed in more than 3/4 of the bed nets surveyed regardless of age, and overall a half of the bed nets were categorized as ‘ineffective’ based on criteria of pHI developed here. The rapidity with which bed nets deteriorated physically was such that less than half of the bed nets were ‘effective’ following two years of use. The results suggest that bed nets are optimally utilized when they are new and physically intact. Bed net utilization decreases gradually with increasing physical deterioration, and most net owners withdraw physically damaged nets from routine use usually after 1.5 years of use [33]. The premise behind consistent use of bed nets during the first year of acquisition and progressively reduced consistency of use in subsequent years is based on their perceived usefulness [18, 34]. Besides being perceived as having diminished mosquito-repelling properties [34], torn nets are also perceived as unlikely to act as a barrier for host-seeking mosquitoes. Thus, the rapid decline in physical condition is almost entirely a function of regular and consistent use, and net owners likely do not find the nets of much use following 1.5 years of use. The finding that most of the ‘effective nets’ were less than two years old and that most of nets not in routine use were more than two years old further support this observation. As demonstrated by Ngufor and others [11], non-use of even the slightly torn nets is likely to become more widespread given growing insecticide resistance of mosquitoes [35]; further diminishing the effectiveness of bed nets. These data imply that net users do not perceive them as protective following two years of use, and it is thus recommended that programmes aim to replace nets on a bi-annual basis.

The absence of a consistent method to measure the physical integrity of bed nets hampers effective comparisons of the findings of this study with those of previous studies. However, a consensus is emerging that under operational conditions physical deterioration significantly shortens bed nets’ effective lifespan from the previously estimated 2–4 years [3–5, 7] to 1.5-2.5 years. Moreover, recent accounts of high proportions of damaged nets within a year of distribution [3, 6, 17, 19, 20] concur with findings that most of nets are physically damaged by 1.5 years of use. Additionally, 1/3 of nets found discarded in a study in Ethiopia were 12–24 months old [34]. These observations help explain the occurrence of substantial number of ‘damaged’ and ‘misused’ nets within the houses surveyed. The majority of nets reported as ‘misused nets’ were actually physically damaged nets. Deployment of these nets to uses other than protection against mosquitoes did not therefore constitute net misuse; the extent of bed net misuse was likely overestimated. This finding re-affirms a recent opinion that bed net misuse is not as widespread as earlier portrayed [36]. Besides physical damage, socio-cultural and environmental factors contributed to withdrawal of nets from routine use (non-utilization) resulting in ‘extra’ and ‘acceptability’ nets. Existence of undamaged nets that are no longer in regular use because they are purportedly ‘extra nets’ or ‘acceptability nets’ points to persistence of some of the barriers to use of bed nets [28]. Intensive information campaigns targeting all individual net users would help lower these barriers.

It is expected that higher bed net use frequency will result in poor physical integrity. The association between reported bed net use and physical condition though positive, was not as strong as anticipated. This inconsistency is likely a consequence of reliance on self-reports to ascribe bed net use. Self-reported bed net use is unreliable because it often misrepresents regular bed net use. Several studies have demonstrated that perceived or actual mosquito density is the main motivation for using bed nets [37–40], and that, therefore, mosquito densities could indirectly be used to predict actual bed net use. In recent mosquito surveys, a general trend of decreasing vector densities with increasing altitude in the study area was observed (Mutuku FM, unpublished data). In three of the four ecological settings with an altitude range of 4–238 meters, the highest proportion of malaria vectors (78%) were collected in villages (Jego and Kidomaya) located at lowest altitude whereas 18% and 4% of the vectors were collected in villages located at the mid (Milalani and Nganja) and highest (Kinango) altitudes respectively. A similar trend in mosquito numbers was observed within the coastal slope villages; where 56%, 38% and 6% of the vectors were collected from the village located at the bottom (Magodzoni), mid (Vuga) and top (Golini) of the slope, respectively. The emerging pattern in the study area of declining mosquito densities with increasing altitude had previously been reported in several other ecological settings [41–43]. Understandably, bed net use decreased with increasing altitude as did the physical deterioration of the nets. Thus, the differences in bed net physical deterioration in the different ecological settings are likely explained by variations in actual bed net use rather than the reported use.

Before ITNs became widely adopted as the primary malaria prevention tool, published data had demonstrated that physically damaged untreated nets were less effective or completely useless in protecting people sleeping under them [44–46]. On the other hand, significantly low protection was offered by torn insecticide treated bed nets; the degree of protection was reported to decrease with increasing physical damage and decreasing insecticide content [4, 5]. Arguably, the low uptake of bed nets then and the perception that torn insecticide treated nets were still effective did not provide the requisite impetus to study the importance of physical damage on effectiveness of insecticide treated nets. However, the recent massive bed net scale-up efforts across Africa and the subsequent adoption of bed nets as the primary malaria prevention tool have created demand for evaluative studies. In addition to reporting impressive bed nets ownership with some countries reaching saturation levels, all these evaluative studies have also reported at least some aspect of physical condition of the nets [11, 16–19, 25, 47, 48]. Despite this current focus, standard procedures for measuring bed net physical integrity are still deficient [2]. Research to elucidate socioeconomic and cultural importance of damaged nets is necessary. How do net owners define a worn-out net or a useless net in as far as protecting oneself against mosquitoes is concerned? How are worn-out nets disposed off and are there uses for worn-out nets? Answers to these questions will provide valuable insight for assessing the useful life span of bed nets.

Differences in net use by demographic category are quite common in many countries in Africa, where higher proportions of pregnant women and infants and in some cases young children (< 6 years) are reported to use nets more commonly than older children (6–18 years old) and other adults [19, 21, 33, 49, 50]. Thus, the finding that bed nets in better physical condition were more likely to be used by infants and pregnant women was not surprising, because the nets are largely acquired free of charge. However, not only are the younger and older children unlikely to use any net but they also use nets of poor physical quality; an observation also reported from western Kenya [20] and Tanzania [50]. These variations in net use and fabric integrity among the different users are mainly attributable to net distribution strategies. It is likely that the vulnerable groups (pregnant women and infants) pass on physically damaged nets to the other non-targeted family members (other adults and older children) [51]. Thus, the well established “catch-up” net distribution strategy in the study area was successful in sustaining high coverage with ‘effective nets’ in pregnant women and infants and as would be expected performed poorly in providing ‘effective nets’ to other non-target groups. These findings reinforce the need for universal coverage with sufficient nets distributed to all household members irrespective of age and malaria vulnerability status. Supplementation of the “catch-up” net distribution strategy with universal mass distribution of nets (“keep-up” strategy) at least every two years would sustain high ownership and utilization of ‘effective nets’ to all net users in the study area.

Bed net maintenance behaviour with regard to net washing, retreatment and repair was unsatisfactory in the study community. Even though only a small proportion of nets found in circulation required insecticide re-treatment, the confusion elicited by the survey questions suggests a large knowledge gap with regard to net retreatment. The threat of non-compliance with net retreatment is minimal because nets that require retreatment are slowly being replaced by LLINs. However, frequent washing of bed nets is associated with loss of insecticide [52], and therefore the finding that about a third of the nets were washed more frequently than recommended merits attention. Furthermore, the data showed that frequent washing often leads to poor physical quality of nets and that light coloured nets were frequently washed relative to dark coloured nets. Malaria vector control programmes are encouraged to primarily procure dark coloured bed nets because they are washed less frequently; moreover non-white nets are widely acceptable in Kenya [28]. In contrast to findings from a study conducted in western Kenya where effective net repairing was reported [53], net repairing in this study area was minimal and its impact in improving physical condition of the nets was minor, given that most repaired nets were categorized as ‘ineffective nets’. Panter-Brick and others [48] working in the Gambia detailed that incomplete repairs (not repairing every single hole on the net), poor quality net fabrics, location of holes (repairs were not done if they were located in parts of the net that could stay tucked under the mattresses), cost of repairing (if a tailor is necessary-for large holes) and time constraints as some the obstacles for effective net repairing [48]. To enhance compliance, it is recommended that instructions on maintenance and washing of nets be delivered both orally and in writing at the time of net acquisition, and that this information be subsequently reinforced regularly, preferably by trained community volunteers through locally appropriate instruction [54]. National malaria control programmes should take note of the increasing importance of bed net’s physical integrity and restructure the behaviour change communication (BCC) strategy to highlight bed net maintenance.

This study was conceived and data were collected before the publication of guidelines for monitoring the durability of LLINs under operational conditions [9]. All the same, much of the methodology used to determine net integrity is similar to the one described in these WHO guidelines [9]. Like any other retrospective cross-sectional survey, this study has a number of potential limitations. Firstly, the outcome of whether nets were used and by whom, their age, frequency of washing and the number of users per net were all based on self-reporting, which is subject to respondent bias. Secondly, since this was a cross-sectional survey conducted in an area where nets of different brands are continuously distributed through numerous delivery channels by multiple groups to the study community, it was impractical to estimate the net survivorship and attrition rates. While the aim of the current study was to document all nets that were currently in routine use, it is likely that some nets were missed, and this may affect the findings. Thirdly, it was not possible to assess the contribution of most net brands to the physical deterioration of nets given that most available nets were of the same colour, net fabric, size, and shape (‘olyset’ brand). It is possible that these net attributes could independently influence the physical condition of bed nets in situations where other distribution channels are employed. Fourth, the exact contribution by each demographic group to physical deterioration of nets may have been affected by within-household factors such bed net sharing, alternating bed net user(s) and sleeping space (mattress or mat), and location (bed room, kitchen or sitting room) [55]. Substantial variation in sharing of nets by the different demographic groups was noted in this study (data not shown). Finally, the definition of ‘ineffective nets’ as used here was somewhat arbitrary, and is only supported by few descriptive studies; hence further research that will lead to an operational definition of ineffective nets is needed.

The results of this study have important implications for malaria vector control programmes using ITNs/LLINs. First, even with proper maintenance and repairs, the results suggest that bed nets should be replaced at least bi-annually. Second, while the “catch-up” bed net distribution strategies are sufficient in ensuring adequate coverage with ‘effective nets ‘in the targeted groups, bi-annual mass distribution is necessary to provide similar coverage to the groups not targeted by “catch-up” strategies. Thirdly, malaria control programmes need to put in place monitoring and maintenance strategies that will deliver locally appropriate education messages on net washing and repair.