The efficacy and cost-effectiveness of insecticide-treated nets (ITNs) in reducing malaria related morbidity and mortality is well-known [1,2] and in recent years has led to massive efforts to distribute millions of free or highly subsidized ITNs to vulnerable populations in sub-Saharan Africa [3-6]. In 2005, the Roll Back Malaria Partnership (RBM) set the goal for 2010 to achieve 80% coverage of children <5 years of age and pregnant women with protective measures such as ITNs [7]. More recently, the Global Malaria Action Plan called for rapid scale-up to universal population coverage for all people at risk for malaria [8]. This scale-up in ITN delivery necessitates an equivalent increase in monitoring and evaluation (M&E) efforts in order to determine the impact of ITN distributions as well as prioritize future programmes.

Typical M&E strategies report on indicators agreed upon and formalized by the RBM Monitoring and Evaluation Reference Group (MERG). Two important core ITN indicators for malaria control programmes are the proportion of households owning an ITN and the proportion of vulnerable populations sleeping under an ITN. Using these measures, many studies have shown that efforts to increase ITN ownership have made tremendous progress; however, most sub-Saharan countries remain well below RBM targets for ownership [9]. In addition, these evaluations have consistently found ITN use (vulnerable groups sleeping under an ITN) lower than household ownership [10-12]. In order to reach the RBM goals for bed net coverage and increase ITN use, it will be useful for programmes to identify and address the determinants of ITN use and non-use.

The two core RBM bed net indicators provide good measures of ITN ownership and use, but in and of themselves cannot provide measures for non-use. The difference between household ITN ownership and individual use cannot be used to estimate non-use because these indicators have different denominators (households and persons, respectively). As a result, there exists literature investigating determinants of bed net use and non-use with a variety of different methods. One method is to restrict analyses to households that already own an ITN in order to observe factors of use other than access [12,13]. For example, in households owning ITNs, children < 5 and women of reproductive age are more likely to sleep under ITNs than other household members, perhaps confirming the effectiveness of messaging that often accompanies distribution or marketing campaigns targeting these high risk groups [14,15]. Data on persons who did not use an ITN the previous night may prove equally informative. A bed net study in Kenya looked at reasons given for not hanging a net and factors that increase or decrease the likelihood that a net will be used; the most important reason for non-adherence was disruption of sleeping arrangements [16].

Following mass distribution campaigns some malaria control programmes implement intervention strategies aimed at increasing ITN coverage. Information, education, and communication/behaviour change communication (IEC/BCC) interventions primarily target and encourage people to hang and sleep under existing bed nets. Hang-up campaigns, send volunteers to selected homes and teach the occupants the importance of sleeping under an ITN, demonstrate how to hang a net, and hang one or more of the household's nets. Keep-up distributions, social marketing, subsidized nets, and distribution through other public health services such as antenatal care clinics aim to improve access to ITNs. These strategies for increasing ITN coverage can be loosely classified into three groups: 1) those aimed at improving ITN access, 2) those encouraging hanging ITNs, and 3) those targeting individuals to encourage sleeping under an existing ITN. Data obtained from surveys evaluating the core RBM bednet indicators can be used to determine what proportion of individuals currently not sleeping under an ITN would be targeted by each of these three categories. Results identifying specific areas of non-use will enable malaria programmes to make informed decisions about which intervention strategies may be more cost-effective to improve ITN coverage.

Although population based indicators of ITN non-use could indeed be programmatically useful, none currently exist. This paper presents a simple framework that supplements information provided by the two core RBM bed net indicators through applying an easy-to-use method to evaluate the different categories of non-use. This technique can be easily applied to data commonly collected on national household cluster survey instruments such as Malaria Indicator Surveys (MIS) and Demographic and Health Surveys (DHS). To demonstrate, this approach was applied to data from household cluster surveys designed to evaluate long-lasting insecticidal net (LLIN) coverage following integrated child health campaigns in five countries.

Response rates were fairly high with 90.8% (Kenya), 86.9% (Niger), 91.7% (Togo), 92.7% (Sierra Leone) and, 92.8% (Madagascar) of the households available for interview.

While many sub-Saharan African countries have made great progress in ITN scale-up, much work remains. An accurate assessment to identify the gaps in household ownership and individual use is a crucial step in devising evidence-based and country-specific strategies to increase population coverage with ITNs and work towards the interruption of malaria transmission. The framework described here, which uses a common denominator (the individual) to evaluate categories of ITN non-use in addition to evaluating the core RBM indicators of household ITN ownership and individual use, provides an additional tool to help accomplish this mission.

In all five countries included in this analysis, the majority of children <5 years of age slept under an ITN the night before the interview. Among children < 5 years not sleeping under an ITN, the largest proportion lived in households that did not own an ITN, despite the efforts of recent child health campaigns. This result, that the largest category of non-use is directly related to household ITN ownership, indicates access is still a barrier to ITN use. This highlights the need to identify distribution strategies targeting previously unreached households to increase ownership of ITNs.

Children <5 years of age living in households hanging an ITN, but not sleeping under one made up a relatively small proportion of non-use for the countries analysed here, suggesting that caretakers recognize the importance of protecting young children with ITNs. In this case, increasing IEC/BCC messages that promote protection of children < 5 with an ITN may not be the most effective method of increasing overall use since they would only be aimed at a small proportion of children not using ITNs. For countries with the smallest proportion of non-use in this category, it may be better to focus efforts on improving one of the other categories of non-use such as increasing ownership or hanging nets. For countries with a large proportion of individuals in this category, further analysis may be needed to determine if there are sufficient numbers of nets per household to cover the population, indicating the need for IEC/BCC to encourage net use, or if there are insufficient nets available, indicating the need for further distribution. In Madagascar and Sierra Leone, additional questions on the determinants of use found the primary reasons for not sleeping under an ITN among persons in households with an ITN hanging, were related to access to an ITN. This suggests further bed net distribution may be more effective than IEC/BCC messages to increase ITN use.

In Niger, Madagascar, and Sierra Leone children <5 years of age living in households owning, but not hanging an ITN, composed a small proportion of the three categories of non-use; suggesting that further hang-up efforts would not lead to major gains in use. In contrast, in Togo and Kenya, a larger proportion of the target population lived in households that owned but did not hang an ITN, indicating a potential gain might be obtained from further a hang-up campaign, other hang-up activities, or increased IEC/BCC programs to encourage net hanging. Although this type of analysis does not address the effectiveness of hang-up activities related to the campaign, it may help identify where to focus efforts and activities beyond those associated with the campaign (by detecting the largest category of non-use) in a cost effective manner.

Countries with a substantial presence of untreated nets (including LLINs that have surpassed their 2-4 year life expectancy) may want to include untreated nets in their evaluation in order to determine their impact on ITN use in households owning both net types. Use of untreated nets may result from differences in behaviour or household dynamics and therefore require interventions that emphasize the advantages of ITNs (or replacing old LLINs) and encourage their use over untreated nets. Alternatively, it could be a result of a shortage of ITNs in the household requiring interventions that aim to increase the number of ITNs owned. Per RBM and WHO recommendations, programme activities should be directed towards increasing access to ITNs for all household members (ie. universal coverage).

The integrated campaigns targeted to children <5 years of age in Madagascar and Sierra Leone detected lower use by other age groups. Mathematical models have shown that targeting vulnerable groups does not necessarily result in adequate total population coverage to achieve a community effect (35-65% of the population sleeping under an ITN) [22]. Results presented here from these two countries suggest targeted campaigns may approach levels of population coverage needed to see a community effect (39.9% and 60.4%). More recently, many national malaria control programmes set a goal of universal coverage, but a standard definition and measurement tool has yet to be determined. Some definitions used for programmatic purposes include: two to three ITNs per household, one ITN for every two people, or one for every sleeping space in a household in an at-risk community. One measure, the proportion of all persons at risk sleeping under an ITN the previous night, was presented in this analysis.

The pattern of ITN use across age groups follow similar trends as published elsewhere [14,15]. Children 5-15 years were most likely to live in a household with an ITN hanging and yet not be sleeping under one. One possible explanation is that, consistent with IEC/BCC messages targeted to at-risk communities, the available ITNs are being used for younger siblings and/or pregnant women/mothers [23]. Another explanation may be the sleeping patterns and functional organization of the household [23,24]. Distribution of LLINs through schools may be a means to reach this population of older children [15]. On a positive note, many of the households had an ITN hanging, even if the child was not under it, which may provide some benefit due to repellent and insecticidal properties. Adults > 48 years were least likely to live in a household with an ITN and current approaches to ITN distribution may need to be modified (e.g. house-to-house methods) to more effectively reach this group and achieve universal coverage.

This analysis provides a useful platform to illustrate a common error made when interpreting M&E results. Using the standard RBM indicators, it may be tempting to consider the deficit in ITN use as the difference between the proportions of household ownership and individual use; however, this would be incorrect because they are calculated using two different denominators. For example, in Niger 65.1% of households owned an ITN and 55.5% of children < 5 years of age slept under an ITN. One may mistakenly conclude the 9.6% difference accounts for the deficit in use. Instead, using this framework one would make the more fitting conclusion that 22% of children < 5 years of age live in households owning an ITN but did not sleep under an ITN (16.4% in households with an ITN hanging and 5.6% not hanging). By classifying individuals into mutually exclusive groups using a common denominator, one is able to appropriately explore the underlying differences between ownership and use.

Likewise, one should use caution when drawing conclusions about the percentage of ITN use from ownership percentages. For example, only 60% of households owning an ITN does not imply a maximum of 60% of children under 5 are sleeping under an ITN. The use/ownership relationship depends on the distribution of children among households owning or not owning nets. Since a household with young children likely has a woman of reproductive age and may be more likely to have other young children as well, it is possible for usage to exceed ownership for certain age groups especially if children share the same sleeping spaces.

Also of note, the definition of a household may greatly affect ownership and reported use and should be considered carefully when comparing national surveys. Countries using definitions that include extended families or polygamous marriages may have a high probability of owning 'at least one ITN' due to the increased number of chances of obtaining a net through free distributions with each woman of reproductive age. However, they also have more household members and as a result each individual may be less likely to use a net because there are not enough nets available for everyone in the household. In contrast, if a household is defined as a mother and her children, then ownership levels may appear lower, but among households owning nets, an individual may be more likely to use it. These subtle differences in the definition of a household highlight a few of the programmatic considerations when interpreting the core indicators of household ITN ownership and individual use and challenge the move toward universal coverage.

The method presented here has important programmatic implications. First, similar to the common RBM indicators, this approach uses simple proportions to evaluate ITN use and does not require performing a logistic regression or other more sophisticated statistical analyses. Second, this approach can be performed using the data currently collected in the common national household surveys such as the MIS and DHS, without the need for additional questions. Lastly, results from this type of analysis can be used to guide programmatic strategies in terms of targeting deficits in ITN use. Population-based surveys can be analysed with this tool and help malaria control programmes select strategies that will help improve ITN use.

There are some potential limitations to the results presented in this paper. First, the survey data used in this analysis were not originally powered for subgroup analysis, resulting in some estimates having wide confidence intervals. Second, the timing of the surveys six to nine months after the campaigns may have introduced information/recall bias if respondents had difficulty remembering events from the campaign, misclassified the net type, or over-reported net use on the basis of social desirability. Third, households unavailable for an interview after multiple attempts or who refused to participate may introduce a non-response bias. Fourth, this paper did not look at the proportion of sleeping spaces covered, which could impact conclusions drawn about both adequate ownership and use. Lastly, although the current analysis looked at all members of the household, future studies could perform analyses accounting for household size and number of ITNs per household or per person, number of sleeping spaces covered; or investigate reasons for non-use among the specific categories described here [16].

The core RBM bed net indicators, household ITN ownership and ITN use by vulnerable groups, play an important role in monitoring ITN distribution programmes. The framework described here supplements these primary indicators and provides an appropriate approach for investigating deficits in ITN use. Using data already available for the primary indicators, one can easily classify individuals into categories of ITN use and non-use with a common denominator (all individuals) and directly determine the distribution of persons among three types of non-use: not owning, not hanging, or not sleeping under an ITN. The framework proposed here uses information commonly collected from national household surveys (DHS, MIS), and as a result it can be readily incorporated into existing ITN monitoring and evaluation strategies to provide information critical to developing tailored activities aimed at improving ITN use.

The authors declare that they have no competing interests.

JVE contributed to the study design, participated in the acquisition and interpretation of data, performed the statistical analysis, and helped draft the manuscript. JT contributed to the design of the study, participated in the acquisition and interpretation of data, and helped draft the manuscript. MAK and AW contributed to the acquisition and interpretation of data, and helped review the manuscript. MA, ME, AH, and LS participated in the concept of the study and coordination, and provided critical review of the manuscript. All authors read and approved the final manuscript.

The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention