Large donations of free nets have allowed net ownership by households in Africa to increase markedly since 2000 [1], and ownership of long-lasting insecticidal nets (LLIN) in Ethiopia has increased dramatically in the last few years [2,3]. After an initial peak in ownership and use shortly after distribution, net ownership and use have been observed to drop off in several countries [4,5].

Clearly, net ownership is a necessary prerequisite for net use. However, whether or not a net owner will use a net every night, some nights, or not at all depends on complex multi-level interactions between individual characteristics, household characteristics, social and cultural factors, community-level factors, aspects of the physical environment and characteristics of the net itself.

In Ethiopia, a decline was observed in net use in households owning nets between two representative household surveys conducted approximately one year apart, and this did not appear to be associated with differences in sampling or any climatic or seasonal differences between the survey years [3]. To investigate further the reasons for the decline, characteristics of nets that may be impacting their use in Ethiopia were studied. Between 2006 and 2007, the proportion of households owning at least one net increased dramatically from 19.6% to 65.6%, but the proportion of nets used the previous night in households owning nets decreased from 85.1% to 56.0% [3,6]. In the net level analysis, factors independently associated in both surveys with reduced likelihood that a net would be used were: increasing net age, increasing damage of nets, increasing household net density (nets/person), and increasing altitude (> 2,000 m). Factors associated with increased likelihood of a net being used were: increasing wealth index (at both surveys), LLIN net type (in 2006), and household status of indoor residual spraying with insecticide (in 2007) [6].

At the individual level, factors influencing net use have been reported to include age and gender [5,7-10], education, occupation/livelihood [11-13], degree of control over household decision-making [14], malaria knowledge, beliefs and risk perceptions [5,15-21], perceived benefits and disadvantages of nets [5,17,18,21], trust in health workers providing health education and LLINs [22], knowledge of appropriate net use/care practices, and net-hanging skills [23-25]. Household level determinants of net use include household size and composition [14], the number of children under five years of age (U5) in the household, intra-household sleeping arrangements [5,13,26], household structure and space [5,7,9,18,26,27], household decision-making processes and power structures [14], and use of other vector control measures [14,19,28]. At the community level, social norms and values [13], cultural beliefs and practices [13,20], mechanisms of LLIN distribution and distance to LLIN suppliers [14,20], rumours about LLINs [20] and social support and pressure [24] all have the potential to influence net use by individuals and within households. For example, white nets may be associated with burial shrouds and death, and free nets may be regarded as toxic or even deliberately harmful to recipient groups [20]. Important environmental factors include climate and temperature [7,18], perceived mosquito density [19], availability and proximity of land for farming and grazing livestock [13]. Characteristics of the nets themselves, such as their cost, size, shape, colour, physical condition, type of insecticide used and perceived durability have also been shown to influence net use [7,19,20], and are likely to interact with individual, household, community and environmental factors in complex ways to determine attitudes towards net use and the feasibility of net use for a given individual or household.

A recent review of literature on determinants of net use highlights the need for greater understanding of these determinants and the relationships between them [29]. In addition, net use among those who own nets is commonly interrupted by temporary, periodic or infrequent conditions, which can inhibit net use even among regular net users. These conditions include travelling, night work, sleeping in the fields during planting or harvest seasons or while tending livestock, attending late-night social events, disruption of usual sleeping arrangements, net unavailability due to washing or dirtiness, extreme fatigue, labour pains, illness or forgetfulness [13,27,30-32].

This study builds on a previous analysis at the net level [6] of the results of two sequential surveys in Ethiopia [3] that demonstrated certain modifiable factors concerning net use, such as improved net care and replacement. In the previous study, the outcome was use or non-use for each net [6], whereas here factors associated with the outcome of individual use or non-use were investigated. The current study adjusts for both household and net level factors, including the important net age and condition characteristics identified previously [6]. In contrast to most previous studies, including some in Ethiopia [19,33,34], this analysis was not restricted to high risk groups (children U5 and pregnant women), but examined net use in all age groups including the 5-24 year old school-age children and young adults, women of reproductive age, and adults age 50 years and older. School-age children are a group generally least protected by insecticide-impregnated nets in Africa [35], including Ethiopia [9].

Since availability of a net in the household is a prerequisite for use, only those households owning at least one net are included in this analysis, to avoid biasing the results by lack of net availability in some households. Net density (i.e. number of nets per person in each household) is also used to account for differing household sizes and intra-household access to nets. The goal of this study is to identify under-served groups and modifiable factors that could be used to better target efforts to increase net use and, hence, assist in long-term reduction of malaria transmission in Ethiopia.

This paper describes determinants of net use as assessed in two consecutive household surveys in 2006 and 2007 in the three largest regional states of Ethiopia. Decline in net use (in households owning nets) between these surveys had been observed, despite increase in overall net ownership [3]. A previous study examined use in a net level analysis (i.e. whether or not each net had been used the previous night) [6], and demonstrated that increasing net age and increasing damage of nets were both associated with a lower likelihood of nets being used. In that study, increased net density was associated with decreased likelihood of a net being used, which is logical since the more nets there are in a household (especially above a density of one net per person), the less the chance of each one being used.

Most previous studies (with some exceptions [19,40]) have expressed net use not as a proportion of nets used, but as a proportion of persons (usually children U5 or pregnant women) using nets [33,34,37,41]. This individual outcome of net use is used in the current study, but includes all individuals in the sampled households since the programme goal is universal coverage. GLLAMM was used to account for both household characteristics and the previously identified net level determinants of use. Using individual net use as an outcome, it would be expected that increased net density would be associated with increased net use, as we found here in both 2006 and 2007. Net density by household was classified in three categories as < 0.5, 0.5 - < 1 and ≥ 1 nets per person, and net use was found to be five-fold higher (in 2006) or two-fold higher (in 2007) when net density was greater than 0.5. This suggests that a net distribution target higher than one net per two persons (such as two nets per three persons or if possible one net per person) is more likely to maximize net use, especially if there are sufficient nets to cover all sleeping spaces. While it is commonly stated that lack of access to sufficient nets within households may be an explanation for low use, there are very few other studies that examined net use in relation to household net density while adjusting for other factors [8].

This study confirmed the previously demonstrated impact of the two related factors of net age and condition on likelihood of net use. In 2006, net use was about twice as likely if some or all the nets in the household were undamaged, while in 2007 net use was one third as likely if all the nets in the household were older than 12 months. The frequent observation of nets in poor condition has been previously reported from Ethiopia [42] and nets owned for less than 12 months were significantly more likely to be used [19]. These findings have implications for care and replacement of nets. The programme should promote net repair and maintenance and also conduct educational activities to change the perception that nets with a few holes are no longer worth using.

Overall, females in both surveys slept under a net the previous night 1.3-1.4 times as often as males. Less than 2% of the population reported being pregnant at each survey. In univariate analysis, pregnant women were at least three times more likely than the general population to use nets, but this factor did not remain associated with net use in the multi-variate model. Net use by age group showed surprisingly that children U5 are not the group with highest net use: persons older than 24 years were more likely to use nets in both surveys, although the association was only statistically significant for the 25-49 year old age group in the 2006 survey. It is clear that children and young adults age 5-24 years use nets much less frequently than children U5 or persons over 24 years (Figure 2). Less frequent use by school-age children has been observed in many other African counties [35] and it would be interesting to conduct a study of net use in which children's own school attendance, access to nets, and knowledge are taken into account, rather than using proxies from women in their households. It is critical that more attention is paid to increasing net use in children, teenagers and young adults (e.g. perhaps by school-based education campaigns and activities), since they likely contribute significantly to transmission and are at risk of severe disease, if infected.

The finding of higher net use in urban compared to rural areas in MIS 2007 was surprising, since urban dwellers might be expected to associate net use with villagers and rural life. However, another study in Ethiopia [19] observed the same higher use of LLIN in urban areas. Possible reasons include: higher levels of education; potentially larger and more diverse social networks perhaps due to greater population density; or a more progressive attitude among urban dwellers leading to earlier adoption of unfamiliar strategies. Other potential unmeasured factors (other than those for which we adjusted such as net density, socio-economic status, malaria knowledge, and IRS status) should be explored further.

Women's malaria knowledge and school attendance were assessed only in the MIS 3R 2007 survey, and expressed as terciles of an index of correct scores, applied at the household level. The results shown here support those of Hwang et al. [15] and others [33,34] that increased women's knowledge of malaria can improve net use by individuals and/or members of their household. However, the interaction effect previously described [15], whereby higher women's knowledge (score of > = 1) only increased net use in themselves and their children U5 if the women had not attended school, was not observed here. This may be due to the different outcome (net use in all ages), different quantification of knowledge scores (in terciles), or application of women's knowledge scores and school attendance at the household level.

Potential limitations of this study include the fact that a number of other possible determinants of net use were not assessed, including, but not limited to, perceptions of malaria risk, the proportion of nets that were hanging, and the opinions or beliefs of householders about their ability to hang nets and/or difficulty of hanging nets in their homes. In addition, aspects of net condition other than the presence of holes (e.g. cleanliness, smell and perceived effectiveness of insecticide) that might affect a person's choice to sleep under a net were not measured. It may be that once the novelty of newly distributed nets has worn off, people stop using nets because the practice grows tiresome and their risk perception returns to initial lower levels, suggesting the need for sustained behaviour change communication. Differences in climate factors or seasonality between the two surveys are potential limitations, but a previous study examined this question extensively [3] and showed that slight differences in survey timing in each year did not result in important differences in rainfall or temperature conditions between the surveys. Differences in sampling strategy between surveys were also examined and were found to be highly unlikely to have introduced bias in the estimate of net use [3].

Examining net use from both the net level [6] and the individual level (present study) while restricting the data to only households owning nets enables adjustment for the different levels of analysis (net, household, cluster) and avoidance of spurious associations, while using net density in the models allows accounting for changes in net ownership over time. Increasing availability of nets within households (i.e. increasing net density), and improving net condition while focusing on education and promotion of net use, especially in school-age children and young adults in rural areas, are crucial areas for intervention to ensure maximum net use and consequent reduction of malaria transmission.

Despite large distribution of new nets between the surveys, the pool of nets in households in 2007 was older and in less good condition than 2006 due to nets remaining from earlier distributions as well as rapid accumulation of damage to nets [43]. Older nets in poor condition are used less than new nets [5]. The study raises important issues about the longevity of nets and when a net is considered 'expired' or no longer suitable for use. More research on this issue is warranted since firstly nets considered to be too damaged may not be used; and secondly the accumulation of 'expired' nets in households would inflate estimates of net ownership in standard household surveys and net density estimates as used in this analysis. Qualitative research investigating when households would consider nets unfit for use, and determining whether this expectation would be realistic - from an operational and programmatic perspective - is necessary. Why damaged and unused nets are kept, or how they are disposed of is also of interest. Such information would generate realistic assessments of average usable net life and improve planning of sensible and sustainable replacement and communication strategies instead of blanket mass distributions at predetermined intervals.

The authors declare that they have no competing interests.

PG, JH, RR, FR, TG and PE conceived and planned the study; PG, JH, TG, AM, ES, ZT, AW and RR trained staff, carried out the survey and managed the data; JN and PG undertook data analysis with input from JH, AM, and RR; PG, JN and AP wrote the paper with input from JH, AM, RR, FR, TG, ZT and PE; all authors approved the final draft.