Fabric integrity, net type and risk index predict attrition of insecticide-treated nets: evidence from a community durability trial of pyrethroid-piperonyl butoxide (PBO) nets in Tanzania Authors: Nolvin John Mvungi, Olukayode G. Odufuwa, Philipo David Ruzige, Joseph B. Muganga, Dorin Joachim Mmasi, Jane Johnson Machange, Dismas S. Kamande, Edith Peter Madumla, Jason D. Moore, Mwema Felix Mwema, Sarah Jane Moore, Emmanuel Mbuba Published: 2025 ABSTRACT BackgroundThe fabric integrity of insecticide-treated nets (ITNs) deteriorates over time during household use, often leading households to discard or stop using damaged nets. This study quantified the extent to which net loss from households is attributable to declining fabric integrity, accounting for variation in user behaviour and environmental conditions.MethodsThis secondary analysis used data from an ITN durability trial in Bagamoyo, Tanzania. A total of 2,514 ITNs (three pyrethroid-PBO and one pyrethroid-only) distributed among 1,509 households were included because they had the outcome of interest (attrition). Fabric integrity was classified as good, damaged, or too-torn according to WHO ITN guidelines and assessed at 12-month intervals. The association between fabric integrity and net attrition over the subsequent 12 months was evaluated using an adjusted logistic regression model, accounting for net type, household risk index, household size, and age of the household head as fixed effects, with household as a random effect.ResultsAttrition increased over time, from 20% at one year to 53% at two years, and 83% at three years post-distribution, and it was associated with net type, household risk index, household size, and age of the household head. Damaged ITNs at one year were more likely to be lost by year two. Compared to nets in good condition, damaged nets had higher odds of attrition (OR: 1.73, 95% CI 1.06–2.82), and too-torn nets were at greater risk (OR: 9.82, 95% CI 5.87–16.4). Similar patterns were observed between years two and three (damaged: OR = 1.55, 95% CI 0.83–2.90; too-torn: OR = 11.7, 95% CI 6.16–22.3).ConclusionFabric integrity was the main driver of rapid ITN attrition and the strongest predictor of loss. Damaged nets were approximately twice as likely to be discarded after one year and at least ten times more likely to be lost in years two and three. These findings highlight the importance of promoting proper net care practices and procuring durable nets that withstand harsh household conditions, ensuring sustained population coverage.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12936-025-05704-z. Background Insecticide-treated nets (ITNs) remain the most effective and widely used malaria intervention globally [1]. Evidence shows that over two-thirds of all malaria cases averted in Africa between 2000 and 2015 were due to ITNs alone [2]. ITNs achieved these reductions by providing both personal and community protection against malaria. Personal protection refers to the direct, individual-level reduction in mosquito bites and therefore disease risk—experienced by a person sleeping under a treated net, which is due to the physical barrier provided by the net [3] as well as sublethal effects of the insecticide, which can prevent mosquito feeding even when the net is damaged [4]. Community protection refers to the indirect reduction in malaria transmission risk experienced by the entire population living in a community where nets have been delivered—including people who do not use nets—as a result of high ITN coverage in a community. Widespread use of ITNs both reduces mosquito population size and average age, reducing their numbers and their capacity to transmit malaria [5–8]. For optimal malaria protection, including benefits to non-users, ITN use among the population at risk of malaria needs to be maximised [9]. In malaria-endemic countries, ITNs are routinely distributed via channels such as antenatal care clinics, the extended programme of immunization and the school net programme every three to four years as part of national strategies to replenish damaged and lost nets [1]. Over 3.1 billion ITNs have been distributed globally between 2004 and 2023 with more than 2.7 (87%) billion ITNs distributed in sub-Saharan Africa (sSA) alone. However, despite this large-scale distribution of ITNs in sSA, the proportion of households with one ITN for every two members was only 45% in 2023. This was largely contributed by the rapid attrition of ITNs from the household that outpaces routine replacements [1]. Attrition refers to the loss of ITNs from the households, which occurs due to discarding from physical damage, giving away to others, selling, theft, or repurposing for non-malaria-related uses [10]. Attrition results from a complex interplay between users’ sociocultural factors, usage environments and ITN characteristics (Fig. 1). These factors not only influence ITN survival but also play a critical role when designing interventions to reduce ITN damage and loss from the household. ITNs tend to last longer when used in households with a female head, with adult users, smaller household size, higher socioeconomic status, improved housing and stronger net care and repair practices [11–14]. Conversely, activities such as cooking or storing food in sleeping areas, drying nets in bushes, not folding the net away during the day, using nets over unfinished beds, the presence of rodents within the household, and crowding under a single net are associated with increased physical damage and faster attrition [13, 15–17]. ITN characteristics such as material type, knitting pattern, and denier also influence attrition due to their influence on fabric durability (damage) [18]. Fig. 1Triad of factors influencing the physical durability of ITNs Triad of factors influencing the physical durability of ITNs However, although damage plays a central role in attrition, its magnitude remains inadequately quantified, particularly among pyrethroid-PBO ITNs, which accounted for more than half of the ITNs distributed in SSA in 2023 [1]. Addressing this evidence gap is essential to inform procurement strategies and optimize the timing of net replenishment. This study aimed to investigate the association between fabric integrity and subsequent attrition of pyrethroid-PBO ITNs. Unlike other studies which often assess ITNs durability by comparing textile quality across brands, this analysis integrates two additional major predictors of physical durability (use environments and user behaviour) (Fig. 1), to construct a risk index [19] for predicting household ITN survival over subsequent years. Methods Study area This study was conducted in the Bagamoyo district (Fig. 2) of the coastal region, located 65 km north of Dar es Salaam, a major commercial city in Tanzania. According to the 2022 Tanzania National Census, the district has a total population of 205,478 people (101,827 males and 103,651 females) [20]. Bagamoyo experiences a warm and humid tropical climate, with annual temperatures ranging between 22 °C and 33 °C, and annual rainfall between 800 and 1,200 mm [21]. Most houses in the district are constructed with brick walls, cement floors, and corrugated iron sheet roofs. The majority of residents engage in small-scale farming of cash crops such as pineapples, rice, coconuts, cashew nuts, and maize [20]. The area has heterogeneous malaria transmission, with prevalence rates ranging from 9 to 30% [22]. ITN access-defined as the proportion of households with at least one ITN for every two potential users is around 69% [23]. Fig. 2Geographical location of Bagamoyo district, Tanzania Geographical location of Bagamoyo district, Tanzania Study design This was a secondary analysis of a prospective household randomized durability trial conducted in Kiwangwa village, Bagamoyo district, Tanzania, between 2020 and 2023 [24]. The primary study aimed to evaluate the physical durability (fabric integrity and attrition) and insecticidal durability (bio-efficacy and residual insecticide) of three pyrethroid-PBO ITNs (herein referred to as pyrethroid-PBO#1, pyrethroid-PBO#2, and pyrethroid-PBO#3) and one pyrethroid-only ITN (PY) over three years of household use [24]. All study ITNs were made of the polyethylene materials, with the same size, colour and shape [24]. Before distribution, all study nets were labelled with a unique identification number printed on self-laminating labels, which are resistant to washing. At baseline, a household survey was conducted to collect sociodemographic information, including household size, house characteristics, and the geographical location of each household. On the same day, study nets were randomly distributed to consenting households. ITN distribution followed a lottery method, in which a field interviewer randomly selected one product from a bag containing all four net types packed in identical cloth bags. Once a product was selected, additional nets of the same type (identified by the same starting digit code) were provided to cover all sleeping spaces in the household. All existing nets in enrolled households were removed to ensure exclusive use of the provided study ITNs. All households within the village, except guest houses, boarding schools, temporary farm dwellings, and those that declined to provide consent, were considered eligible for the primary study. Follow-up surveys were conducted on selected ITNs at one-year intervals to assess attrition, fabric integrity, bio-efficacy, and residual chemical content, as per the WHO guidelines [25]. Data on fabric integrity and attrition were the only data used for this secondary analysis study. Data extraction Data from the primary study were received in Stata format. Key information extracted included ITN fabric integrity, attrition, household socioeconomic status, household size, attitudes toward net care and repair, and net-handling practices (such as drying location, daytime hanging, and cooking or food storage in the sleeping room). Additional data on the usage environment were also extracted, including the presence of rodents in the sleeping room, wall materials, bed type, type of fuel used for cooking, and the gender, education, and age of the household head. Data were collected at 12-, 24-, and 36 months post-distribution using a structured questionnaire [26]. At each survey, the presence or absence of study ITNs was recorded, with reasons documented for any missing nets. For each net found in the household, holes of different sizes were counted following WHO guidelines [27]. In this study, each hole size category was assigned a weight: 1, 23, 196, and 576 for sizes 1–4. The proportionate hole index (pHI) was calculated as;\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pHI = \left( {{1 } \times {\text{ number of small holes}}} \right) \, + \, \left( {{23 } \times {\text{ number of medium holes}}} \right) \, + \, \left( {{196 } \times {\text{ number of large holes}}} \right) \, + \, \left( {{576 } \times {\text{number of very large holes}}} \right).$$\end{document}pHI=1×number of small holes+23×number of medium holes+196×number of large holes+576×number of very large holes. Study outcome The primary outcome was attrition, defined as the proportion of initially distributed ITNs no longer present in the household. It was calculated by dividing the number of ITNs lost due to wear and tear, or repurposing by the total number of study nets distributed at baseline. This was also referred to as Type 1 attrition or true attrition. ITNs lost due to theft, selling, use elsewhere, or given away were considered as lost to follow-up (Type 2 attrition), and were not included in the regression analysis. Primary exposure variable The primary exposure variable was fabric integrity, defined as good (pHI < 64), damaged (pHI 64–642), or too-torn (pHI > 642). Data management Data from the baseline, 12-, 24-, and 36-month surveys were merged using the ITN unique identification number to create one dataset. Nets with incomplete data were removed from the dataset. Data management involved identifying and removing duplicates, renaming variables and labels, and generating new variables as needed. Key variables generated for the final analysis included socio-economic status (SES), net care and repair attitude, and household risk index (RI) score. The Principal Component Analyses (PCA), were conducted following the approach outlined by Vyas et al. [28], to generate the SES variable for each household. Variables that were indicative of ownership of durable assets (radio, mobile phone, fan, television, iron, and refrigerator), farm animals (chickens, ducks, goats, and cows), house structure (wall, roof, and floor), means of transportation (bicycle, motorcycle, tricycle, and car), and access to utilities (water, toilet), and a light source (electricity, battery/solar panel, and hurricane lamp) were used to generate the SES variable. The SES variable was categorized into three groups: low, middle, and high. A Likert scale used by Helinski et al. [29] was used to estimate the net care and repair attitude of the respondents using a set of seven questions on a scale of 2 to −2; strongly agree, somewhat agree, somewhat disagree, and strongly disagree. These questions asked: (1) whether mosquito nets are valuable, (2) whether they can take actions to make nets last long, (3) whether it was possible to repair holes in nets, (4) whether repaired nets can still be effective against mosquitoes (5) or if they have time to repair a hole in their nets (6) or whether they can protect their family from malaria by taking care of the net and (7) if they are confident they can repair holes in their nets immediately. The average of these variables was used to obtain a score categorized into two: poor/average attitude (< 1), and positive attitude (≥ 1). These questions have been refined and are part of the ITNs durability monitoring questionnaire [26]. A risk index score used by Kilian and Koenker [19] that aggregated variables related to net care and repair attitudes, handling practices, and usage environments was used to reduce the number of variables in a dataset that could impact the complexity of analysis. A higher household risk index score refers to a higher risk of damage and subsequently attrition for the ITNs in that household. To estimate the risk index score, the variables were organized into three thematic groups, each weighted according to its contribution to the composite score (Additional file 1). The first category, representing net handling practices, contributed 45% to the total risk index score. It included variables such as food storage in the sleeping area, cooking inside the sleeping room, net drying practices, and the condition of the net during the day. The second category, comprising environmental factors, contributed 10% to the total risk index score. It included variables such as wall structure, presence of rodents in the household, type of cooking fuel, and bed type. The third category, also weighted at 45%, included net care and repair attitudes and whether participants recalled any care or repair of their nets. Scores from all three categories were summed and then classified into two groups: lower risk (score ≤ 50) and higher risk (score > 50). Data analysis Analyses were conducted in STATA 18.0 (StataCorp, College Station, TX, USA). Descriptive statistics were performed using means, frequencies, and percentages. The logistic regression analysis was conducted for each net followed-up in the study that was not lost to follow-up. Therefore, attrition was a binary outcome (0 = net present or 1 = net lost through type 1 attrition). Covariates were first evaluated in univariable analyses and were included in the multivariable model if their p-value was < 0.25 [30]. Three separate mixed-effect logistic regression models were performed to compare net performance across different time points: (1) 12 versus 24 months, (2) 24 versus 36 months, and (3) 12 versus 36 months. All models included fabric integrity (primary covariate), net type, wealth index, household risk index, household size, and head of household characteristics (gender and age) as fixed effects and household as a random effect. Model fit was tested using Hosmer–Lemeshow goodness of fit test. Collinearity was assessed with no sign of unacceptable correlation detected. The final model used adjusted odds ratio (aOR) with 95% CI, considering p-values < 0.05 as statistically significant to reject the null hypothesis that there was no association between fabric integrity and attrition. Results Baseline characteristics of the households Over 13,500 ITNs distributed, a total of 2,514 ITNs (625 pyrethroid-PBO#1, 637 pyrethroid-PBO#2, 590 pyrethroid-PBO#3, and 662 pyrethroid-only nets) from 1,509 households were included in this analysis. The study arms had similar sociodemographic and household characteristics (Table 1). Most houses had mud walls, cement floors, and corrugated iron roofs. The average household size and number of sleeping spaces were similar across study arms, and the majority (> 70%) of households were male-headed (Table 1). Table 1Baseline sociodemographic and household characteristics of the study area per study armCharacteristicPyrethroid-only Pyrethroid-PBO#1Pyrethroid-PBO#2Pyrethroid-PBO#3Mean number of sleeping spaces per HH2.92.72.82.7Mean number of people per HH5.55.25.15.0n (%)n (%)n (%)n (%)Number of ITNs662 (26)625 (25)637(25)590 (24)Gender of the head of the HH Male513 (77)466 (75)464 (73)452 (77) Female149 (23)159 (25)173 (27)138 (23)Age of head in years 18–34172 (26)176 (28)184 (29)211 (36) 35–59205 (31)216 (35)198 (31)214 (36) 60–99285 (43)233 (37)255 (40)165 (28)Education level of the head None154 (23)141 (22)149 (23)110 (19) Primary433 (65)411 (66)434 (68)391 (66) Secondary or higher75 (12)76 (12)54 (9)89 (15)Socioeconomic status Low241 (36)257 (41)266 (42)154 (26) Middle258 (39)198 (32)224 (35)170 (29) High163 (25)170 (27)147 (23)266 (45)Household size 1–3357 (54)369 (59)383 (60)361 (61) 4 and above305 (46)256 (41)254 (40)229 (39)Wall Mud wall461 (70)464 (74)447 (70)360 (61) Bricks201 (30)161 (26)190 (30)230 (39)Floor Earth292 (44)276 (44)278 (44)188 (32) Cement370 (56)349 (56)359 (56)402 (68)Roof Thatch30 (5)25 (4)33 (5)17 (3) Iron sheets632 (95)600 (96)604 (95)573 (97) Table 1Baseline sociodemographic and household characteristics of the study area per study armCharacteristicPyrethroid-only Pyrethroid-PBO#1Pyrethroid-PBO#2Pyrethroid-PBO#3Mean number of sleeping spaces per HH2.92.72.82.7Mean number of people per HH5.55.25.15.0n (%)n (%)n (%)n (%)Number of ITNs662 (26)625 (25)637(25)590 (24)Gender of the head of the HH Male513 (77)466 (75)464 (73)452 (77) Female149 (23)159 (25)173 (27)138 (23)Age of head in years 18–34172 (26)176 (28)184 (29)211 (36) 35–59205 (31)216 (35)198 (31)214 (36) 60–99285 (43)233 (37)255 (40)165 (28)Education level of the head None154 (23)141 (22)149 (23)110 (19) Primary433 (65)411 (66)434 (68)391 (66) Secondary or higher75 (12)76 (12)54 (9)89 (15)Socioeconomic status Low241 (36)257 (41)266 (42)154 (26) Middle258 (39)198 (32)224 (35)170 (29) High163 (25)170 (27)147 (23)266 (45)Household size 1–3357 (54)369 (59)383 (60)361 (61) 4 and above305 (46)256 (41)254 (40)229 (39)Wall Mud wall461 (70)464 (74)447 (70)360 (61) Bricks201 (30)161 (26)190 (30)230 (39)Floor Earth292 (44)276 (44)278 (44)188 (32) Cement370 (56)349 (56)359 (56)402 (68)Roof Thatch30 (5)25 (4)33 (5)17 (3) Iron sheets632 (95)600 (96)604 (95)573 (97) Baseline sociodemographic and household characteristics of the study area per study arm Attrition Of all ITNs selected for this analysis, 20% (n=494) were lost 12 months post distribution. This increased to 53% (n = 1,319) at 24 months and further to 83% (n = 2,083) at 36 months. The causes of net loss varied over time, with lost to follow-up (i.e., use elsewhere, sale, theft, or being given away) being the primary cause of net loss within the first 12 months, accounting for 11% (n = 275) of all net losses. After the first 12 months, discarding due to wear and tear became the leading cause of attrition, accounting for 37% (n = 923) of all nets that were lost by 24 months and 63% (n = 1,574) of all nets lost by 36 months (Table 2). Table 2Proportion of attrition by causes across the survey periodsN = 2,51412 months24 months36 monthsn%n%n%Type 1 attrition (true attrition)2199923371,57463Type 2 attrition (lost to follow-up)275113961650920All cause attrition (cumulative losses)494201,319532,08383Remained2,020801,1954743117Percentages were calculated by dividing number of ITNs lost across the surveys (n) by the total number of ITNs at baseline (N) Table 2Proportion of attrition by causes across the survey periodsN = 2,51412 months24 months36 monthsn%n%n%Type 1 attrition (true attrition)2199923371,57463Type 2 attrition (lost to follow-up)275113961650920All cause attrition (cumulative losses)494201,319532,08383Remained2,020801,1954743117Percentages were calculated by dividing number of ITNs lost across the surveys (n) by the total number of ITNs at baseline (N) Proportion of attrition by causes across the survey periods Percentages were calculated by dividing number of ITNs lost across the surveys (n) by the total number of ITNs at baseline (N) Fabric integrity Out of the 2,514 ITNs included in this analysis, 2,020 (80%) were still in household possession at 12 months, 1,195 (47%) at 24 months, and 431 (17%) at 36 months (Table 2). The proportion of ITNs in good physical condition declined steadily, particularly after 12 months post-distribution, while those in too-torn condition increased. Compared to the pyrethroid-only net, pyrethroid-PBO ITNs had a faster rate of deterioration from good to too-torn physical condition (Fig. 3). Fig. 3Lost ITNs and the fabric integrity of those remained within household per arm per survey Lost ITNs and the fabric integrity of those remained within household per arm per survey Univariate analyses of factors associated with attrition Higher attrition at 24 months and 36 months was associated with decreasing fabric integrity in the previous survey, with the highest attrition among too-torn ITNs compared to those in good condition. Attrition also varied across net brands, being significantly higher among pyrethroid-PBO#3, and least among the pyrethroid-only ITN. Attrition increased with higher wealth status, larger household size, and higher household risk index. In contrast, increasing age of head of the household and female gender were associated with lower attrition. However, there was no association between education level of head of the household and attrition. Overall, attrition was independently associated with fabric integrity, net type, SES, risk index, household size, gender, and age of head of the household (Table 3). Table 3Univariable analyses between attrition and other predictor variablesPredictor name24 months36 monthsOR (95% CI)P-valueOR (95% CI)P-valueFabric integrity Good11 Damaged1.49 (1.17–1.91)0.0011.36 (1.00–1.85)0.053 Too-torn5.43 (4.39–6.72)< 0.0016.85 (5.19–9.04)< 0.001Net type Pyrethroid net11 Pyrethroid-PBO#11.55 (1.20–2.01)0.0011.57 (1.22–2.03)0.001 Pyrethroid-PBO#22.97 (2.31–3.82)< 0.0013.07 (2.30–4.09)< 0.001 Pyrethroid-PBO#33.53 (2.72–4.57)< 0.0016.82 (4.71–9.88)< 0.001Socioeconomic status Low11 Middle1.46 (1.19–1.80)< 0.0011.36 (1.07–1.73)0.012 High1.79 (1.44–2.22)< 0.0011.53 (1.18–1.97)0.001Household size Small11 Large1.57 (1.32–1.87)< 0.0011.55 (1.26–1.93)< 0.001Household risk index Low11 High2.09 (1.75–2.49)< 0.0011.95 (1.57–2.42)< 0.001Gender of household head Male11 Female1.00 (0.81–1.22)0.9660.71 (0.57–0.89)0.003Age of household head Young adults (18-34)11 Adults (35-59)1.02 (0.82–1.27)0.8810.74 (0.55–1.00)0.053 Older adults (60-99)0.56 (0.45–0.70)< 0.0010.34 (0.26–0.44)< 0.001ORL: Odds ratio, CI: confidence interval. Values under the 24-month columns represent the odds of attrition at 24 months based on fabric integrity at 12 months, while values under the 36-month columns represent the odds of attrition at 36 months based on fabric integrity at 24 months. All other covariates were measured at baseline Table 3Univariable analyses between attrition and other predictor variablesPredictor name24 months36 monthsOR (95% CI)P-valueOR (95% CI)P-valueFabric integrity Good11 Damaged1.49 (1.17–1.91)0.0011.36 (1.00–1.85)0.053 Too-torn5.43 (4.39–6.72)< 0.0016.85 (5.19–9.04)< 0.001Net type Pyrethroid net11 Pyrethroid-PBO#11.55 (1.20–2.01)0.0011.57 (1.22–2.03)0.001 Pyrethroid-PBO#22.97 (2.31–3.82)< 0.0013.07 (2.30–4.09)< 0.001 Pyrethroid-PBO#33.53 (2.72–4.57)< 0.0016.82 (4.71–9.88)< 0.001Socioeconomic status Low11 Middle1.46 (1.19–1.80)< 0.0011.36 (1.07–1.73)0.012 High1.79 (1.44–2.22)< 0.0011.53 (1.18–1.97)0.001Household size Small11 Large1.57 (1.32–1.87)< 0.0011.55 (1.26–1.93)< 0.001Household risk index Low11 High2.09 (1.75–2.49)< 0.0011.95 (1.57–2.42)< 0.001Gender of household head Male11 Female1.00 (0.81–1.22)0.9660.71 (0.57–0.89)0.003Age of household head Young adults (18-34)11 Adults (35-59)1.02 (0.82–1.27)0.8810.74 (0.55–1.00)0.053 Older adults (60-99)0.56 (0.45–0.70)< 0.0010.34 (0.26–0.44)< 0.001ORL: Odds ratio, CI: confidence interval. Values under the 24-month columns represent the odds of attrition at 24 months based on fabric integrity at 12 months, while values under the 36-month columns represent the odds of attrition at 36 months based on fabric integrity at 24 months. All other covariates were measured at baseline Univariable analyses between attrition and other predictor variables ORL: Odds ratio, CI: confidence interval. Values under the 24-month columns represent the odds of attrition at 24 months based on fabric integrity at 12 months, while values under the 36-month columns represent the odds of attrition at 36 months based on fabric integrity at 24 months. All other covariates were measured at baseline Multivariate analyses of factors associated with attrition in the subsequent years After controlling for net type, socioeconomic status, household size, household risk index, household clustering, gender, and age of head of the household, there was still a significant association between fabric integrity (damage) and ITN attrition (loss) in the subsequent years (Table 4). The odds of attrition at 24 months given their physical condition at 12 months were 73% higher (OR: 1.73, 95% CI 1.06–2.82 P = 0.027) among damaged ITNs and nine times higher (OR: 9.82, 95% CI 5.87–16.4; P = 0.000) among too-torn ITNs compared to good ITNs. ITN condition in year 1 also predicted their likelihood of being lost by year 3 with similar probabilities as for being lost in year 2. Table 4Logistic regression to measure the association between fabric integrity and attritionPredictor name12 vs 24 months24 vs 36 months12 vs 36 monthsaOR (95% CI)P-valueaOR (95% CI)P-valueaOR (95% CI)P-valueFabric integrity Good111 Damaged1.73 (1.06–2.82)0.0271.55 (0.83–2.90)0.1722.08 (1.27–3.42)0.004 Too-torn9.82 (5.87–16.4)< 0.00111.7 (6.16–22.3) < 0.0016.78 (3.77–12.2) < 0.001Net type Pyrethroid-only net111 Pyrethroid-PBO#11.82 (0.95–3.48)0.0732.18 (1.15–4.10)0.0162.29 (1.19–4.42)0.013 Pyrethroid-PBO#26.88 (3.40–13.9)< 0.0017.45 (3.52–15.8)< 0.0019.34 (4.31–20.3)< 0.001 Pyrethroid-PBO#36.50 (3.07–13.7)< 0.00120.1 (7.9–51.3)< 0.00123.8 (9.1–62.3)< 0.001Socioeconomic status Low111 Middle2.25 (1.30–3.90)0.0041.70 (0.95–3.03)0.0741.87 (1.04–3.38)0.038 High2.71 (1.51–4.84)0.0011.28 (0.69–2.37)0.4381.66 (0.88–3.14)0.115Household size Small111 Large2.17 (1.34–3.53)0.0022.36 (1.38–4.04)0.0022.56 (1.48–4.44)0.001Household risk index Low111 High1.77 (1.13–2.78)0.0122.78 (1.67–4.64)< 0.0011.84 (1.08–3.12)0.025Gender of household head Male111 Female1.31 (0.78–2.23)0.3090.68 (0.39–1.19)0.1770.65 (0.37–1.14)0.131Age of household head Young adults (18-34 years)111 Adults (35-59 years)0.77 (0.43–1.37)0.3690.48 (0.25–0.95)0.0340.42 (0.21–0.83)0.013 Older adults (59-99 years)0.36 (0.20–0.64)< 0.0010.15 (0.07–0.29)< 0.0010.12 (0.06–0.24)< 0.001 Household Identification number9.1 (6.36–13.02)8.34 (5.02–13.9)9.39 (5.7–15.5)The odds ratio (OR) and 95% confidence interval (95% CI) is the estimates from the mixed-effect logistic regression adjusted for net type, household size, household risk index, wealth index, household clustering, gender, and age of head of the household with their 95% confidence interval. Values under the 12 vs 24-month and 12 vs 36-month columns represent the odds of attrition at 24 and 36 months, respectively, based on fabric integrity at 12 months. Values under the 24 vs 36-month columns represent the odds of attrition at 36 months based on fabric integrity at 24 months. All other covariates were measured at baseline Table 4Logistic regression to measure the association between fabric integrity and attritionPredictor name12 vs 24 months24 vs 36 months12 vs 36 monthsaOR (95% CI)P-valueaOR (95% CI)P-valueaOR (95% CI)P-valueFabric integrity Good111 Damaged1.73 (1.06–2.82)0.0271.55 (0.83–2.90)0.1722.08 (1.27–3.42)0.004 Too-torn9.82 (5.87–16.4)< 0.00111.7 (6.16–22.3) < 0.0016.78 (3.77–12.2) < 0.001Net type Pyrethroid-only net111 Pyrethroid-PBO#11.82 (0.95–3.48)0.0732.18 (1.15–4.10)0.0162.29 (1.19–4.42)0.013 Pyrethroid-PBO#26.88 (3.40–13.9)< 0.0017.45 (3.52–15.8)< 0.0019.34 (4.31–20.3)< 0.001 Pyrethroid-PBO#36.50 (3.07–13.7)< 0.00120.1 (7.9–51.3)< 0.00123.8 (9.1–62.3)< 0.001Socioeconomic status Low111 Middle2.25 (1.30–3.90)0.0041.70 (0.95–3.03)0.0741.87 (1.04–3.38)0.038 High2.71 (1.51–4.84)0.0011.28 (0.69–2.37)0.4381.66 (0.88–3.14)0.115Household size Small111 Large2.17 (1.34–3.53)0.0022.36 (1.38–4.04)0.0022.56 (1.48–4.44)0.001Household risk index Low111 High1.77 (1.13–2.78)0.0122.78 (1.67–4.64)< 0.0011.84 (1.08–3.12)0.025Gender of household head Male111 Female1.31 (0.78–2.23)0.3090.68 (0.39–1.19)0.1770.65 (0.37–1.14)0.131Age of household head Young adults (18-34 years)111 Adults (35-59 years)0.77 (0.43–1.37)0.3690.48 (0.25–0.95)0.0340.42 (0.21–0.83)0.013 Older adults (59-99 years)0.36 (0.20–0.64)< 0.0010.15 (0.07–0.29)< 0.0010.12 (0.06–0.24)< 0.001 Household Identification number9.1 (6.36–13.02)8.34 (5.02–13.9)9.39 (5.7–15.5)The odds ratio (OR) and 95% confidence interval (95% CI) is the estimates from the mixed-effect logistic regression adjusted for net type, household size, household risk index, wealth index, household clustering, gender, and age of head of the household with their 95% confidence interval. Values under the 12 vs 24-month and 12 vs 36-month columns represent the odds of attrition at 24 and 36 months, respectively, based on fabric integrity at 12 months. Values under the 24 vs 36-month columns represent the odds of attrition at 36 months based on fabric integrity at 24 months. All other covariates were measured at baseline Logistic regression to measure the association between fabric integrity and attrition The odds ratio (OR) and 95% confidence interval (95% CI) is the estimates from the mixed-effect logistic regression adjusted for net type, household size, household risk index, wealth index, household clustering, gender, and age of head of the household with their 95% confidence interval. Values under the 12 vs 24-month and 12 vs 36-month columns represent the odds of attrition at 24 and 36 months, respectively, based on fabric integrity at 12 months. Values under the 24 vs 36-month columns represent the odds of attrition at 36 months based on fabric integrity at 24 months. All other covariates were measured at baseline The odds of attrition at 36 months given their physical condition at 24 months were 55% higher (OR: 1.55, 95% CI 0.83–2.90; P = 0.172) among damaged ITNs and 11 times higher (OR: 11.7, 95% CI 6.16–22.3; P = 0.000) among too-torn ITNs compared to good ITNs. Overall, attrition increased as fabric integrity declined. Attrition was associated with ITN type, larger household size, higher household socioeconomic status, and higher household risk index. In contrast, attrition decreased with increasing age of head of the household (Table 4). Discussion This study aimed to quantify the magnitude of attrition due to a change in the fabric integrity of ITNs over time in the durability trial of ITNs in Bagamoyo, Tanzania. The study demonstrated that attrition was associated with fabric integrity. In this setting, the fabric integrity of an ITN measured one year earlier was a consistent predictor of attrition in the following year, even after adjusting for user sociodemographic characteristics and the usage environments. These results are similar to another study conducted in Tanzania, which reported that for each additional hole in an ITN, the likelihood of that net being lost increased by about 5% [31]. This study identified loss of fabric integrity (wear and tear) as a strong driver of ITN attrition. Damage-related attrition was low for the first 12 months post-distribution, then in year 2 accounted for 37% of all nets lost by 24 months and 63% by 36 months. Similar patterns have been reported in a multi-country analysis of ITNs, where 63% of nets leaving households were discarded due to physical damage [10]. Two studies conducted in Tanzania’s Lake Zone also highlighted fabric deterioration as a key factor in attrition, contributing up to two-thirds of all ITN loss within three years [32, 33]. By contrast, a study in Afghanistan reported only 12% attrition over 36 months, suggesting differences in environmental, cultural, or operational contexts [34]. These findings suggest that in this setting, the standard three-year ITN distribution cycle may not sustain adequate coverage, warranting context-specific interventions. Such interventions include approaches such as stratified ITN distribution, currently employed by the Tanzania National Malaria Control Programme (NMCP), to increase ITN coverage among the populations most in need [35, 36] and the use of intensified continuous distribution through the use of reproductive and child health channel and school net programme. Nonetheless, production of durable ITNs with a longer functional lifespan should be a central strategy to reduce operational costs associated with frequent replenishment and to minimize environmental impact of discarded nets. Attrition varied by net type, with pyrethroid-PBO#3 nets being more likely to be lost than the other pyrethroid-PBO ITNs and the standard pyrethroid nets. These differences may reflect brand-specific construction characteristics, such as denier, knitting pattern and bursting strength in addition to chemical composition [18, 37, 38]. Although the material type of ITNs is known to influence their physical durability and subsequently attrition [18], it is unlikely to have contributed significantly to the observed differences in this study, as all nets were made of polyethylene material [24]. The effects of denier, knitting and bursting strength can be reflected using the resistance-to-damage (RD) scores [39], which predicted the highest damage among brands with low RD score (low snag strength, low bursting strength and low resistance to hole enlargements) compared to those with high (Mbuba et al., pers. commun.). Regardless, pyrethroid-PBO nets remain effective in reducing malaria incidence even when too torn; however, this protective effect is short-lived, persisting only while the concentration of active ingredients remains sufficiently high [40, 41]. Therefore, efforts to improve the fabric integrity of pyrethroid-PBO must be prioritised, as well as tailored Behaviour Change Communication (BCC) to ensure their care for a longer-lasting effect [29]. In this cohort, more than half of ITNs were lost within the first two years of use, with less than a quarter remaining in the household possession for up to 36 months. Consistent with these findings, several other durability studies in Tanzania have documented that fewer than half of ITNs remain in use by 36 months, largely due to swift loss of fabric integrity [16, 18, 32, 33, 42]. In contrast, a study conducted in Kenya has reported very low attrition of ITNs, with over 90% of the nets surviving to 36 months [43]. This variability can be explained by differences in socio-cultural practices , frequency of ITN use, as well as geographical location, as reported in other African countries such as Nigeria [44] and Ethiopia [45]. Some of these differences may be partially reflected in the risk index, which incorporates environmental, behavioural, and attitudinal factors influencing ITN survival [19]. At 12 months, this study found that Type 2 attrition (11%), was slightly higher than Type 1 attrition (9%). These results are consistent with other studies highlighting the role of user behaviour in ITN survival [13, 44, 46]. Similarly, a multi-country study conducted in Uganda, Ghana, Nigeria and Senegal reported high rates of removal (i.e. Type 2 attrition) soon after distribution, particularly among campaign-distributed ITNs [10]. This trend was also observed in Benin, where removal remained the leading cause of attrition not only within the first 12 months but also up to 24 months of net use [47]. However, contrary to these findings, a study in Ethiopia reported wear and tear as the primary reason for attrition, even within the first 12 months following distribution [17]. In this study, large households had an increased likelihood of their ITNs being lost compared to small households, similar to findings observed in Mozambique and Ethiopia [12, 17]. These results may be due to the high risk of damage among ITNs in larger households, given the high likelihood of overcrowding under a single ITN during sleeping [15, 23]. However, households headed by older heads were less likely to lose their ITNs compared to those headed by young adults, as also observed from a previous study in this setting [31] as well as a multi-country study which involved Nigeria, Tanzania, Mozambique, and DRC [13]. Younger household heads are more likely to have young children, who are strongly associated with ITN damage [15]. The study found that households with high SES had higher attrition rates at 12 months compared to the low SES group. This likely reflects greater ability among wealthier households to replace too-torn ITNs with new ones, as reported in another study [31]. However, this was not consistent at the 24- and 36-month surveys, probably because the initial wave of attritions stabilised the net population in the high SES group. Conversely, other studies have observed lower attrition rates among the highest wealth status households [13]. While gender has been identified as an important predictor of attrition in other studies [14], in this study, no significant association was observed between the gender of the head of the household with attrition. This study has several limitations. Firstly, some of the households' information used in these analyses was collected at baseline, which may have underestimated the effects of their changes over time, such as shifts in attitudes toward net care and repair, or changes in socioeconomic status (SES). Second, the analysis was based on data from a single study conducted at one site, which may limit the generalizability of the findings to the broader national or international context. Third, the use of secondary data may have limited the overall scope of the study, as only nets that were followed up to 36 months were included, and factors such as loss to follow-up were not further explored. The large proportion of attrition due to lost to follow-up (20% at 36 months) could introduce bias if the condition of these removed ITNs systematically differed from those retained, thereby affecting the classification and interpretation of results. These limitations should be considered when interpreting the results. Conclusion The study highlights fabric integrity as a primary predictor of ITN attrition and that early damage to ITNs is likely to cause attrition at a faster rate. Therefore, procurement should prioritize ITNs with demonstrated fabric strength. Environmental and behavioural factors (risk index) play a crucial role in the long-term functional survival of ITNs, and addressing these factors can significantly reduce early net damage. Demographic factors including household size, and the age of the head of the household were also significantly associated with attrition. Overall, these findings highlight the importance of procuring durable nets that withstand harsh household conditions and promoting proper net care practices to maximise ITN durability and cost effectiveness. Supplementary Information Additional file1 Additional file1