Deworming Coverage: What 29 Surveys Say About Equity
Preschool deworming coverage across sub-Saharan Africa came to only 49.2 percent when 29 national surveys were pooled. Treatment concentrated among children from wealthier and better-educated households, and the adjusted gap between the top and bottom wealth fifths reached 13.7 percentage points. Every child treated consumes one dose, which is why such figures feed straight into albendazole tablet forecasts.
Why preschool deworming coverage lags school-age programs
Four worm species carry the soil-transmitted helminth burden: Ascaris lumbricoides, Trichuris trichiura, Necator americanus and Ancylostoma duodenale. Global infection is estimated at more than 1.5 billion people, with 642 million prevalent cases counted in 2021, and sub-Saharan Africa absorbs an unusually large share of both transmission and disability. Young children accumulate the damage: blood lost through the gut, poor micronutrient absorption, iron-deficiency anemia, stunted growth and weaker cognitive development.
The operational answer is preventive chemotherapy. Fixed-interval single doses of albendazole 400 mg or mebendazole 500 mg are the recommended regimen for endemic communities, and the 2021-2030 roadmap asks for 75 percent coverage among preschool- and school-aged children in endemic areas. Cheap and safe as the drugs are, treatment alone does not stop transmission. Community trials of water, sanitation and hygiene improvement found only modest protection against reinfection, with a relative risk of 0.86 and a confidence interval of 0.74 to 1.01 spanning the null.
Where preschool children lose out is in delivery. School-aged children have a centralized platform in mass drug administration run through schools. Children between 12 and 59 months mostly do not attend primary school, so their treatment depends on biannual child health days, vitamin A campaigns or a routine facility visit. Once campaign delivery becomes irregular, the fallback is an opportunistic clinic visit or a pharmacy purchase paid out of pocket, and at that point medication fees, travel distances and caregiver health literacy start to decide the outcome.
How this deworming coverage analysis was assembled
Data came from 29 nationally representative household surveys fielded between 2006 and 2024, pooled as a cross-sectional set. Every survey followed a two-stage stratified cluster design: census enumeration areas drawn with probability proportional to size, then a systematic household sample within each cluster. Living children aged 12-59 months at interview formed the population, with completed months of age reconstructed from the interview date and the birth date on record.
Sampling proceeded through two exclusion rules. Infants aged 0-11 months were dropped, since preventive chemotherapy with albendazole or mebendazole is not recommended below 1 year. Children lacking a usable deworming status were dropped as well. Screening opened with 37 national surveys and 365,650 children under five; eight surveys held no deworming variable, leaving 307,592 enumerated children in the 29 retained surveys. After 64,720 infants and 19,278 children without an outcome were removed, 223,594 children remained for descriptive work and 219,097 entered the regression, a 2.01 percent loss to item non-response.
The outcome variable records whether a caregiver says the child received medicine for intestinal worms during the 6 months preceding the survey. Interviewers carried visual job aids and tablet sample cards, letting caregivers identify albendazole and mebendazole chewable tablets by sight rather than recalling a brand or a dose.
Inequality was computed, not inferred. Within each survey, children were ordered by a continuous household asset score and given weighted midpoint fractional ranks; two summaries followed, a conventional concentration index and an Erreygers-corrected index that rescales for a bounded binary outcome so that surveys with very different coverage levels can be compared. Adjusted odds came from a survey-weighted logistic model carrying country fixed effects and cluster-robust standard errors at the survey-PSU level. Absolute contrasts came from marginal standardization with every other covariate held at its observed distribution, and delta-method intervals derived from the cluster-robust covariance matrix.
Analytic element | Specification |
Population | Living children aged 12-59 months at the interview date |
Surveys pooled | 29 nationally representative household surveys |
Period covered | 2006 to 2024 |
Descriptive sample | 223,594 children |
Regression sample | 219,097 children, 2.01 percent lost to item non-response |
Outcome | Caregiver report of medicine for intestinal worms in the previous 6 months |
Exclusion rule | Infants under 12 months, because preventive chemotherapy is not recommended under 1 year |
Wealth ranking | Continuous household asset score, ranked within each survey |
Inequality metrics | Conventional concentration index and Erreygers-corrected index |
Regression | Survey-weighted logistic model with country fixed effects and cluster-robust standard errors |
Absolute contrasts | Marginal standardization with delta-method confidence intervals |
Pooled deworming coverage against the 75 percent benchmark
Across the 223,594 children, pooled deworming coverage was 49.2 percent (95 percent CI 48.6 to 49.9). Weighting each country equally instead, which ignores sample size, raised the average to 52.5 percent (46.1 to 58.9). That divergence is informative rather than contradictory: the biggest national samples come from settings where coverage runs low.
Country estimates ran from 19.6 percent (17.3 to 22.0) at the floor to 88.9 percent (87.6 to 90.1) at the ceiling. Only three surveys passed the 75 percent marker, at 88.9, 81.1 and 75.9 percent. Eight fell in the 60 to 75 percent band, five dropped under 35 percent, and 26 of 29 remained below the benchmark.
Quoting that gap without context would be unfair to the programs. The 75 percent figure is an operational threshold for preventive chemotherapy inside endemic implementation units, while a national survey counts coverage across endemic and non-endemic districts in the same denominator. The two are indicative of reach; neither settles whether a program is working.
Where deworming coverage divides: wealth and education
Wealth alone produced a clean gradient before any adjustment: 38.8 percent coverage in the poorest fifth, then 44.2, 49.1, 55.0 and 63.3 percent in the richest. Maternal schooling climbed further, from 36.2 percent among children of mothers with no formal education through 51.9 percent at primary and 58.3 percent at secondary to 69.6 percent at higher education. Both gradients cleared p < 0.001.
Three more variables behaved the same way. Urban coverage reached 56.8 percent against 45.3 percent rural. Children of employed mothers reached 53.3 percent against 43.1 percent. Improved drinking water gave 52.1 percent against 42.9 percent, and improved sanitation 54.7 percent against 43.8 percent. Sex was the exception, at 49.4 percent for boys and 49.1 percent for girls with p = 0.281, which is what a gender-blind program should produce.
Dimension | Subgroup | Weighted coverage, percent [95 percent CI] |
All children | Pooled sample | 49.2 [48.6-49.9] |
Household wealth | Poorest | 38.8 [37.8-39.8] |
Household wealth | Poorer | 44.2 [43.2-45.1] |
Household wealth | Middle | 49.1 [48.2-50.1] |
Household wealth | Richer | 55.0 [54.0-56.1] |
Household wealth | Richest | 63.3 [62.2-64.4] |
Maternal education | No education | 36.2 [35.3-37.1] |
Maternal education | Primary | 51.9 [51.1-52.7] |
Maternal education | Secondary | 58.3 [57.4-59.1] |
Maternal education | Higher | 69.6 [68.0-71.2] |
Residence | Urban | 56.8 [55.8-57.8] |
Residence | Rural | 45.3 [44.5-46.0] |
Maternal employment | Employed | 53.3 [52.6-54.0] |
Maternal employment | Not employed | 43.1 [42.3-43.9] |
Drinking water | Improved | 52.1 [51.4-52.8] |
Drinking water | Unimproved | 42.9 [42.0-43.9] |
Sanitation | Improved | 54.7 [53.9-55.5] |
Sanitation | Unimproved | 43.8 [43.0-44.6] |
Child sex | Boys | 49.4 [48.7-50.1] |
Child sex | Girls | 49.1 [48.4-49.8] |
How unequal deworming coverage is across the wealth distribution
Both inequality indices point the same way. The pooled Erreygers index reached 0.2362 (0.2229 to 0.2494), and the conventional concentration index reached 0.1199 (0.1132 to 0.1267). Positive values mean treatment clusters among richer children. The conventional figure is the smaller of the two only because its feasible range contracts as mean coverage approaches either bound, which is exactly the distortion the corrected index removes and the reason cross-survey comparison should rest on the corrected one.
Labels were assigned by significance rather than by arbitrary cutoffs, and 25 of 29 surveys qualified as pro-rich with intervals excluding zero. The sharpest gradients reached 0.4128 (0.3502 to 0.4753) and 0.3935 (0.3593 to 0.4277). A single survey reversed the pattern with significant pro-poor concentration of -0.1106 (-0.1811 to -0.0402). Three produced intervals crossing zero, at 0.0200 (-0.0207 to 0.0607), -0.0381 (-0.0903 to 0.0141) and -0.0532 (-0.1098 to 0.0035).
Even the best-performing survey, at 88.9 percent coverage, retained measurable inequality of 0.0328 (0.0091 to 0.0565). Designers should read that pair as good news with a caveat: near-universal reach drags the index toward zero without closing the residual gap, so coverage volume and equity appear to answer to the same delivery choice.
Index or statistic | Estimate | 95 percent CI | Reading |
Pooled Erreygers index | 0.2362 | 0.2229 to 0.2494 | Pro-rich concentration overall |
Pooled conventional index | 0.1199 | 0.1132 to 0.1267 | Same direction, compressed by coverage level |
Steepest national Erreygers index | 0.4128 | 0.3502 to 0.4753 | Sharpest wealth gradient in the set |
Second steepest national index | 0.3935 | 0.3593 to 0.4277 | Low coverage combined with a steep gradient |
Highest-coverage survey | 0.0328 | 0.0091 to 0.0565 | Inequality small but still significant |
Only pro-poor survey | -0.1106 | -0.1811 to -0.0402 | Significant concentration toward poorer children |
Survey with a null interval | 0.0200 | -0.0207 to 0.0607 | Confidence interval crosses zero |
Survey with a null interval | -0.0381 | -0.0903 to 0.0141 | Confidence interval crosses zero |
Survey with a null interval | -0.0532 | -0.1098 to 0.0035 | Confidence interval crosses zero |
Which factors predict deworming coverage after adjustment
Country fixed effects did not dislodge the two leading correlates. Compared with children of mothers holding no formal education, adjusted odds were 1.43 (1.37 to 1.49) at primary, 1.84 (1.76 to 1.93) at secondary and 2.41 (2.21 to 2.64) at higher education. Moving up the wealth quintiles from the poorest fifth gave 1.18 (1.13 to 1.23), 1.35 (1.28 to 1.42), 1.59 (1.49 to 1.70) and 1.91 (1.76 to 2.07).
Demographic terms held too. Mothers aged 35-49 years carried adjusted odds of 1.38 (1.31 to 1.45) against the 15-24 group. Children aged 24-35 months showed 1.21 (1.17 to 1.25) against 12-23 months, children aged 36-47 months showed 1.13 (1.09 to 1.17), and the oldest band showed nothing distinguishable from the reference at 1.02 (0.99 to 1.06). Birth order declined step by step: 0.99 (0.96 to 1.02) for the second and third child, 0.93 (0.89 to 0.98) for the fourth and fifth, and 0.87 (0.82 to 0.92) from the sixth onward.
Contextual variables added smaller shifts. Employed mothers corresponded to adjusted odds of 1.31 (1.26 to 1.35), improved drinking water to 1.11 (1.06 to 1.16) and improved sanitation to 1.07 (1.03 to 1.11). Sex remained flat at 0.99 (0.97 to 1.02).
Factor | Category | Adjusted OR [95 percent CI] | p |
Maternal education | No education | 1.00 [reference] | Reference |
Maternal education | Primary | 1.43 [1.37-1.49] | < 0.001 |
Maternal education | Secondary | 1.84 [1.76-1.93] | < 0.001 |
Maternal education | Higher | 2.41 [2.21-2.64] | < 0.001 |
Household wealth | Poorest | 1.00 [reference] | Reference |
Household wealth | Poorer | 1.18 [1.13-1.23] | < 0.001 |
Household wealth | Middle | 1.35 [1.28-1.42] | < 0.001 |
Household wealth | Richer | 1.59 [1.49-1.70] | < 0.001 |
Household wealth | Richest | 1.91 [1.76-2.07] | < 0.001 |
Maternal age | 15-24 years | 1.00 [reference] | Reference |
Maternal age | 25-34 years | 1.19 [1.14-1.24] | < 0.001 |
Maternal age | 35-49 years | 1.38 [1.31-1.45] | < 0.001 |
Child age | 12-23 months | 1.00 [reference] | Reference |
Child age | 24-35 months | 1.21 [1.17-1.25] | < 0.001 |
Child age | 36-47 months | 1.13 [1.09-1.17] | < 0.001 |
Child age | 48-59 months | 1.02 [0.99-1.06] | 0.223 |
Birth order | First-born | 1.00 [reference] | Reference |
Birth order | 2-3 | 0.99 [0.96-1.02] | 0.537 |
Birth order | 4-5 | 0.93 [0.89-0.98] | 0.004 |
Birth order | 6 or more | 0.87 [0.82-0.92] | < 0.001 |
Maternal employment | Not employed | 1.00 [reference] | Reference |
Maternal employment | Employed | 1.31 [1.26-1.35] | < 0.001 |
Drinking water | Unimproved | 1.00 [reference] | Reference |
Drinking water | Improved | 1.11 [1.06-1.16] | < 0.001 |
Sanitation | Unimproved | 1.00 [reference] | Reference |
Sanitation | Improved | 1.07 [1.03-1.11] | < 0.001 |
Residence | Urban | 1.00 [reference] | Reference |
Residence | Rural | 0.96 [0.91-1.02] | 0.207 |
Child sex | Male | 1.00 [reference] | Reference |
Child sex | Female | 0.99 [0.97-1.02] | 0.490 |
Absolute gaps matter more than odds ratios
Marginal standardization restates the model in percentages a planner can work with. With every other covariate held at its observed distribution, predicted receipt was 43.1 percent (42.2 to 44.0) for the poorest fifth, then 46.6 (45.8 to 47.3), 49.4 (48.7 to 50.2), 52.9 (52.0 to 53.8) and 56.8 percent (55.6 to 58.0) for the richest. The adjusted wealth gap between those two poles is 13.7 percentage points (12.0 to 15.4).
Schooling stretched the range further. Predicted margins moved from 42.2 percent (41.5 to 42.9) for mothers with no education to 49.8 (49.1 to 50.5), 55.3 (54.6 to 56.0) and 61.0 percent (59.3 to 62.7) at primary, secondary and higher levels, an absolute gap of 18.8 percentage points (17.0 to 20.6).
Holding both metrics in view changes the story. An adjusted odds ratio of 1.91 for the richest fifth reads as a strong effect, yet it maps to a 31.8 percent relative lift in predicted probability and only a 13.7 point absolute difference. Relative measures flatter slow progress, because a large ratio applied to a modest base still leaves many children untreated. Reporting odds alone overstates how close a program stands to equity.
Dimension | Subgroup | Predicted margin, percent [95 percent CI] | Absolute contrast vs reference |
Household wealth | Poorest | 43.1 [42.2-44.0] | Reference |
Household wealth | Poorer | 46.6 [45.8-47.3] | +3.4 pp [2.5-4.4] |
Household wealth | Middle | 49.4 [48.7-50.2] | +6.3 pp [5.2-7.4] |
Household wealth | Richer | 52.9 [52.0-53.8] | +9.8 pp [8.4-11.2] |
Household wealth | Richest | 56.8 [55.6-58.0] | +13.7 pp [12.0-15.4] |
Maternal education | No education | 42.2 [41.5-42.9] | Reference |
Maternal education | Primary | 49.8 [49.1-50.5] | +7.6 pp [6.7-8.5] |
Maternal education | Secondary | 55.3 [54.6-56.0] | +13.1 pp [12.1-14.1] |
Maternal education | Higher | 61.0 [59.3-62.7] | +18.8 pp [17.0-20.6] |
The counterintuitive result: rural residence is not the barrier
The raw comparison suggests a simple answer that adjustment overturns. Urban coverage stood 11.5 percentage points above rural coverage, 56.8 percent against 45.3 percent. Enter wealth, maternal education, water and sanitation, and rural residence loses all independent effect: adjusted odds 0.96 (0.91 to 1.02) with p = 0.207.
Composition explains the gap. Urban households in these samples hold more wealth, more schooling and better infrastructure, and once those are modeled the geographic tag carries no information. The implication for policy is uncomfortable. Targeting rural districts because they look underserved spends money on the wrong variable and leaves poor urban children outside the net entirely.
Here the indices and the regression converge. Household wealth and caregiver schooling drive individual receipt, geography does not, and the concentration index is constructed from precisely that household ranking. Community-level work on socioeconomic barriers covers the rural and the urban poor alike; district-level targeting covers one group only.
What a 49.2 percent deworming coverage rate means for albendazole supply
Before coverage is a health statistic it is a procurement variable. One treatment episode equals one dose, and the recommended regimen is single-dose albendazole 400 mg or mebendazole 500 mg, so planned tablet volume follows eligible children, rounds per year and a wastage allowance. Because treatment is not recommended below 1 year, extending a campaign to the 12-23 month band enlarges the eligible denominator before any order is placed.
Delivery route shapes ordering behavior as much as volume does. Campaign platforms such as child health days compress demand into a narrow window, which favors suppliers able to hold finished stock before the round opens. Passive clinic and pharmacy channels release demand across the year in smaller lots. These questionnaires cannot separate the two: the platform through which a child was treated is unrecorded, as are local stockouts, and the study reports both gaps as limitations.
Equity feeds back into demand less obviously. Where roughly half of eligible children are reached and the unreached half skews toward poor and less-schooled households, the marginal tablet in a well-run program lands in a household that campaign delivery already misses. Hence the case for pairing facility supply with community outreach, and hence the reason program demand can jump when outreach expands while the eligible population stands still.
Upstream, that argues for reading albendazole demand off campaign architecture and coverage targets rather than off population growth. Registration, dossier maintenance and stability work follow the markets where campaigns operate, and the tablet format follows the age band, since chewable presentations are what campaign workers hand to caregivers.
Limitations of the deworming coverage evidence
The figures carry several caveats. A cross-sectional design makes every association a conditional correlate rather than a causal effect. Surveys from 2006 through 2024 span different phases of national worm control, so the pooled number describes a mix rather than one contemporary moment; limiting the model to surveys from 2015 onward gave 1.94 (1.78 to 2.12) for the richest fifth and 2.42 (2.21 to 2.65) for higher maternal education, which leaves both gradients intact.
Recall is a second issue, since receipt depends on a caregiver remembering 6 months back. Neither infection status nor local transmission intensity is measured, so endemicity never enters the adjustment. Household wealth is ranked within each country against national peers, not valued in purchasing power, which makes the index ordinal across surveys at best. Platform and stockout data are absent, so no channel can be credited with the coverage observed. Stratification appeared in the descriptive and inequality work but not in the regression covariance, a compromise the authors disclose, because country fixed effects and survey strata together produce a high-dimensional problem.
Losses were limited but clustered. Among 4,497 children dropped for item non-response, 83.6 percent of exclusions traced to one national survey that omitted floor material. Dropping that covariate moved estimates little: 1.89 (1.76 to 2.04) for the richest fifth and 2.43 (2.22 to 2.65) for higher education. Dropping the largest national sample instead shifted the richest-quintile odds to 1.81 (1.66 to 1.97) and the higher-education odds to 1.97, a reminder of how much one large survey can move a pooled model.
Sensitivity model | Sample | Richest vs poorest aOR | Higher vs no education aOR |
Main model | 219,097 | 1.91 [1.76-2.07] | 2.41 [2.21-2.64] |
Largest national sample removed | 199,048 | 1.81 [1.66-1.97] | 1.97 |
Surveys from 2015 onward | 191,019 | 1.94 [1.78-2.12] | 2.42 [2.21-2.65] |
Parsimonious socioeconomic model | 219,097 | 2.00 | 2.41 |
Floor material omitted | 222,874 | 1.89 [1.76-2.04] | 2.43 [2.22-2.65] |
Bottom line
A pooled deworming coverage figure of 49.2 percent is the headline; the shape of the distribution is the actual finding. Twenty-six of 29 surveys stayed below the 75 percent marker, 25 showed significant pro-rich concentration, and the adjusted gaps of 13.7 percentage points for wealth and 18.8 points for maternal education are wide enough to plan against. Once household capital entered the model, rural residence stopped predicting anything, which shifts the focus from place to household economics and caregiver schooling. Anyone working upstream in the albendazole chain should read that twice: total tablet demand scales with campaign reach, and the form of the residual gap decides whether the next increment of demand arrives as bulk campaign orders or as small facility orders.
FAQ
What is deworming coverage and how is it measured
It is the share of eligible children reported to have received medicine for intestinal worms inside a defined recall window, here the 6 months before the interview. In these surveys the question is asked of the caregiver, and interviewers use visual job aids and tablet sample cards so that albendazole and mebendazole chewable tablets can be recognized without naming a brand. Coverage is then estimated with survey weights, so each national sample represents its own child population rather than only the households that were visited.
How high is deworming coverage among preschool children in sub-Saharan Africa
Pooled coverage across 29 nationally representative household surveys was 49.2 percent (95 percent CI 48.6 to 49.9). The unweighted equal-country average was 52.5 percent (46.1 to 58.9). National estimates ranged from 19.6 percent to 88.9 percent, and 26 of 29 surveys sat below the 75 percent benchmark. Roughly half of surveyed preschool children therefore had no reported treatment in the previous 6 months.
Why is the 75 percent benchmark hard to compare with survey estimates
The benchmark is an operational threshold written for preventive chemotherapy delivered inside endemic implementation units. A national household survey measures coverage across endemic and non-endemic districts together, so its denominator includes children at little or no risk of infection. The two numbers can be placed side by side as an indication of reach, but a survey estimate below 75 percent does not by itself prove that endemic districts are under-served.
Which children are least likely to receive deworming
Children in the poorest households and children of mothers with no formal education. Unadjusted coverage was 38.8 percent in the poorest fifth against 63.3 percent in the richest fifth, and 36.2 percent where the mother had no schooling against 69.6 percent with higher education. Sixth-born or later children were also less likely to be treated, with adjusted odds of 0.87 (0.82 to 0.92) relative to first-born children.
How unequal is deworming coverage across wealth quintiles
The pooled Erreygers index was 0.2362 (95 percent CI 0.2229 to 0.2494) and the conventional concentration index was 0.1199 (0.1132 to 0.1267), both positive and therefore pro-rich. Twenty-five of 29 surveys showed statistically significant pro-rich concentration. One survey showed significant pro-poor concentration at -0.1106, and three had confidence intervals crossing zero, so the pooled figure conceals real national variation.
Which factors remain associated with deworming receipt after adjustment
Maternal education and household wealth stayed strongest, at adjusted odds of 2.41 (2.21 to 2.64) for higher education against none and 1.91 (1.76 to 2.07) for the richest fifth against the poorest. Older maternal age, older child age and maternal employment were positive, while higher birth order was negative at 0.87 (0.82 to 0.92). Improved water and sanitation contributed small positive associations.
Does rural residence still reduce the chance of deworming
No. The crude gap was real, with urban coverage at 56.8 percent against 45.3 percent in rural areas, but rural residence was not independently associated with receipt after adjustment for wealth, maternal education and household infrastructure: adjusted odds 0.96 (0.91 to 1.02), p = 0.207. The urban advantage reflects the concentration of wealthier and better-schooled households in cities rather than geography acting on its own.
How large is the absolute gap between richest and poorest children
About 13.7 percentage points after adjustment. Marginal standardization put predicted coverage at 56.8 percent (55.6 to 58.0) in the richest fifth and 43.1 percent (42.2 to 44.0) in the poorest, a gap of 13.7 points (12.0 to 15.4). The equivalent maternal education gap is wider still, at 18.8 percentage points (17.0 to 20.6) between higher education and no education.
How does maternal education change the probability of receipt
Predicted coverage rose from 42.2 percent (41.5 to 42.9) where the mother had no formal education to 49.8 percent with primary, 55.3 percent with secondary and 61.0 percent (59.3 to 62.7) with higher education. That is an adjusted spread of 18.8 percentage points, wider than the wealth gap in the same model, which makes caregiver schooling the strongest social gradient in the dataset.
What does a 49.2 percent coverage rate mean for albendazole supply
It sets the volume side of the equation. Each treatment episode is one dose, and the recommended regimen is single-dose albendazole 400 mg or mebendazole 500 mg, so campaign tablet requirements scale with eligible children aged 12-59 months, planned rounds and wastage. Because roughly half of eligible children are currently reached, the unreached half is the growth margin, and reaching that margin usually means community outreach rather than more fixed-facility supply.
What are the main limitations of these survey estimates
Four limits matter most. Receipt is caregiver-reported recall over 6 months, so reporting error cannot be excluded. The questionnaires record neither the delivery platform nor local stockouts, which blocks attribution to campaign or clinic channels. Infection status and local transmission intensity are unmeasured, so no adjustment for endemicity is possible. Finally the surveys span 2006 to 2024, which mixes different stages of national control rather than describing a single moment.