azithromycin treatment coverage in tanzanian children using community volunteers
TRANSCRIPT
Azithromycin treatment coverage in Tanzania 167
Correspondence and reprint requests to:Dr. Sheila West, Ph.D.Dana Center Prev. Ophthalmol.Johns Hopkins Univ. Sch. Med.600 N. Wolfe St.Baltimore, MD 21205USA.
Acknowledgements:This project was supported by fundsfrom the Edna McConnell ClarkFoundation and the InternationalTrachoma Initiative. Dr. West is aResearch to Prevent BlindnessSenior Scientific Investigator. Theauthors wish to thank Dr. NealHalsey and Dr. Peter Winch for theirvaluable suggestions on earlier draftsof this paper.
Abstractpurpose To determine which of two village-based strategies wasmore effective at recruiting residents for a trachoma mass treatmentcampaign.
methods The two strategies were to use either village governmentpersonnel to recruit residents for treatment, or to solicit interestedcommunity volunteers to recruit residents.Three were villages assignedto each strategy, and the outcome measured was treatment coverage ofindividuals, group and the villages.
results Self-selected community volunteers were significantly moreeffective than village government personnel in recruiting villagers forantibiotic treatment (p < .0001). The differences were strongest for thegroup at highest risk for active trachoma, pre-school children; 73% ofchildren in community volunteer villages were treated, compared to63% in village government villages (p < .05). Children in villages usingcommunity volunteers and from larger families were more likely to betreated.
conclusion These findings support using motivated community volunteers, rather than traditional government workers, for mass treatment campaigns where high coverage is necessary.
Key words Trachoma treatment; azithromycin; patient recruit-ment; children; Tanzania
Ophthalmic Epidemiology0928-6586/03/US$ 16.00
Ophthalmic Epidemiology– 2003, Vol. 10, No. 3,pp. 167–175© Swets & Zeitlinger 2003
Accepted 17 October 2002
Azithromycin treatment coverage in Tanzanian children using
community volunteers
Matthew Lynch, PhD1
Sheilak West, PhD1
Beatriz Muñoz, MS1
Kevin D. Frick, PhD2
Harran A. Mkocha, BS3
1Dana Center for Preventive Ophthalmology and 2BloombergSchool of Public Health, Johns Hopkins University, Baltimore,MD, USA and 3Kongwa Trachoma Project, Kongwa, Tanzania
Original article
Oph
thal
mic
Epi
dem
iol D
ownl
oade
d fr
om in
form
ahea
lthca
re.c
om b
y U
nive
rsity
of
Con
nect
icut
on
10/2
9/14
For
pers
onal
use
onl
y.
Introduction Trachoma, caused by an ocular infection with C.trachomatis, is the second leading cause of blindness worldwide. Thesafety and ease of the one-dose azithromycin antibiotic regimen offersan attractive means of community antibiotic treatment for trachomacontrol.Azithromycin is generally well-tolerated, with mild side-effects,and a single dose is sufficient for clearance of the chlamydial infection.1
The efficacy of azithromycin in community level treatment was shownin a 1994 study, in which intensively supervised research teams im-plementing three rounds of treatment were able to cover 90% of thepopulation with at least one dose, producing a 64–93% reduction in chlamydial ocular infection.2
The World Health Organization, in conjunction with an azithromycindonation program from Pfizer, Inc., has established a global initiativefor trachoma control in endemic countries.3 Optimal trachoma controlrequires a community-level approach4,5 and should include componentsfor surgery, antibiotic treatment, facial cleanliness and environmentalimprovement (the SAFE strategy).6
Community-directed treatment programs in which local communitiestake charge of recruitment and treatment procedures have been shownto improve compliance for onchocerciasis control with ivermectin.7
Working through traditional chiefs or other village government leadersto organize the treatment process in each village has led to success-ful interventions. A program in Uganda also found that community volunteers selected by “community leaders” achieved high rates (64%in the first year) of compliance using a community-based distributormodel.8 However, using community leaders, either to recruit house-holds or to select volunteers, while convenient for program planners, islikely to depend on underlying communication networks and informalgroups within the village. Groups outside leadership social networksmay be missed.
Households in communities can be recruited to treatment throughmedia information campaigns or inter-personal contacts. Oladepo et al.9 showed that either an information media or a social networkapproach using existing community groups could significantly increaserecruitment for facility-based ivermectin treatment in Nigeria,although the social network approach using community groups wassubstantially cheaper.
The effect of alternative recruitment strategies for implementingcommunity-wide treatment with azithromycin has not been evaluated.Our study compares coverage rates for two community treatmentschemes using inter-personal contact recruitment of households: viaself-selected community volunteers (CV) versus via elected village government (VG) personnel in a trachoma-endemic area of centralTanzania.
Methods
project site The project was located in Kongwa district of centralTanzania.This is a rural area with large villages (1500–8000 population)consisting of small central cores and large peripheral areas with morescattered households of farmers and herders. Trachoma prevalence
168 M. Lynch et al.
Oph
thal
mic
Epi
dem
iol D
ownl
oade
d fr
om in
form
ahea
lthca
re.c
om b
y U
nive
rsity
of
Con
nect
icut
on
10/2
9/14
For
pers
onal
use
onl
y.
averages 60% among pre-school children aged 1–8.10 Villages in Tanzania are divided into geographic neighborhoods (mtaa) which inturn are divided into smaller ten-cell units headed by an elected rep-resentative (balozi). The project conducted a complete census of all villages prior to beginning work.
study design
Village enrollment This study supplemented an ongoing project com-paring treatment via school-based screening versus via householdrecruitment by village government personnel. Three villages, identifiedon the basis of anticipated high rates of trachoma and access withinone hour’s travel of the project office in Kongwa, were randomly allo-cated to a strategy with household recruitment implemented throughthe village governments (VG villages). For this additional study ofhousehold recruitment through community volunteers (CV), threeadditional villages were selected from a village list using the same cri-teria, but selected on the basis of smallest village sizes, as drug supplieswere limited for this supplemental study. The limitations imposed bythis non-random selection are discussed below.
In the VG villages the village government, via the balozi leaders, wasresponsible for recruiting households with pre-school children fortreatment. In the CV villages, this function was carried out by self-selected volunteers, recruited by the project team in the course of aweek’s visit prior to start of the treatment campaign. Team membersrecruited volunteers in the course of interviews, casual discussions andmeetings with local groups such as churches. Any village governmentpersonnel who volunteered were not excluded in the CV villages,though the project team reported that the numbers were small.
Seminars were held in each village with either the village governmentpersonnel (VG arm) or the community volunteers in which participantsplanned the treatment campaign, including recruitment procedures.Planning procedures were the same in each village. In the CV villages,volunteers were asked to bring a small quantity of maize meal to theseminar to contribute to the mid-day meal, as a means of demonstrat-ing interest. This was the only difference between the organizationalseminars in the two arms.
Treatment implementation In both the VG and CV villages, allmembers of households with pre-school children (aged 6 mos-7 years)were targeted for treatment (eligible population). Recruitment wasdone through face-to-face contact by individuals going door to door,explaining the treatment and encouraging participation by households.No posters or other media tools were used.
Treatment procedures Treatment was carried out by the Kongwa Trachoma Project team, with assistance from volunteers from theabove-mentioned groups at a designated sites in each neighborhood of each village. Treatment was offered free of charge to all members of target households without examination, and any communitymember requesting examination with clinical signs of trachoma
Azithromycin treatment coverage in Tanzania 169
Oph
thal
mic
Epi
dem
iol D
ownl
oade
d fr
om in
form
ahea
lthca
re.c
om b
y U
nive
rsity
of
Con
nect
icut
on
10/2
9/14
For
pers
onal
use
onl
y.
indicating active disease. The team was available for one day in eachneighborhood, approximately one week per village. Treatment con-sisted of azithromycin in water suspension (approximately 20mg/kg,range 16–23mg/kg) for persons weighing less than 50kg, and 1000mgin 500mg tablet form for persons weighing 50kg and more. Pregnantwomen were screened for signs of clinical trachoma prior to treatmentand those with active trachoma given azithromycin. Pregnant womenwith no signs of active trachoma were treated with topical tetracycline1% ointment. An eye nurse was responsible for collecting data on anyadverse reactions in the village at the time of treatment. To documentcoverage a project team member identified each person receiving treat-ment from project census lists and recorded her/his study ID numberon the treatment form, then observed the drug being taken. Follow-upof non-attenders was the responsibility of those in charge of recruit-ment (village government personnel or community volunteers) whoused their own knowledge of people in their neighborhood.
Research and treatment protocols were reviewed and approved bythe Tanzania National Prevention of Blindness Committee, the Tanza-nia National Institute for Medical Research and the Johns HopkinsHospital Joint Committee on Clinical Investigation.
survey methods A questionnaire survey of between 70–100 house-holds randomly selected from among 400 households with pre-schoolchildren in each village was conducted one week prior to the treatmentto collect more detailed information on predictors of compliance withtreatment. At this pre-treatment visit, all household members underage 15 were offered free eye examinations, in which trachoma statuswas graded for severity using the WHO Simplified Grading Scheme.11
The same households were re-visited two weeks after the treatmentwith a follow-up questionnaire. Data from these questionnaires are notpresented here, but selection for the survey was included in the modelpredicting treatment compliance.
data management and analysis Treatment and interview formswere entered into computer databases using customized data entryprograms in Paradox 4.0 (Borland, Inc.). All treatment data enteredwere printed out in summary form and checked against the originalforms to ensure accuracy.
For analysis at the individual level, three groups at varying risks ofactive trachoma were identified based on the known epidemiology oftrachoma. The highest risk group was pre-school children and thelowest was adult males. Coverage in each arm was defined as thenumber of individuals in each risk group from target households whoreceived treatment divided by the total number of individuals in thatrisk group from target households.
Coverage data for each risk group were first organized into contin-gency tables using SAS (SAS Corp., Cary, NC) and crude relationshipsexamined with chi-square analysis. Factors postulated as important orsignificant in bivariate analysis were included in logistic regressionmodels. Models included terms for VG versus CV recruitment arms,and having been selected for the questionnaire survey. Other factors
170 M. Lynch et al.
Oph
thal
mic
Epi
dem
iol D
ownl
oade
d fr
om in
form
ahea
lthca
re.c
om b
y U
nive
rsity
of
Con
nect
icut
on
10/2
9/14
For
pers
onal
use
onl
y.
table 1. Characteristics of villagesby recruitment group.
included were number of children in the household, and whether aschool-aged child was reported to be attending school. These factorswere posited to reflect the potential difficulty of getting multiple chil-dren organized and to a central site and attitudes towards community-level institutions (schools). Generalized Estimating Equation (GEE)analysis was used to adjust for the effect of intra-household clusteringof treatment coverage.
Results VG villages were larger, had more children per eligiblehousehold, and the residents were more likely to report sending theirschool-aged children to school (Table 1). The random sample of house-holds selected for the questionnaire survey was not significantly dif-ferent from the pool of eligible households in each village (data notshown). Due to smaller village sizes in the CV arm, the sample com-prised a larger proportion of the eligible households in the CV group(~29%) compared to the VG group (~18%).
The proportion of the eligible population (all members of householdswith children <8 years old) treated in each set of villages was 51.6% inthe VG group and 58.6% in the CV group. Treatment coverage variedby the age and sex-defined risk groups (Table 2). The percent of eachgroup treated was highest among pre-school children, declined forwomen and school-aged females, and dropped sharply for school-agedmales and men. For each risk group, coverage rates of eligible personswere higher for CV villages than for villages in the VG group.
The strongest predictor of treatment for pre-school children was thetype of recruitment, with children in villages using community volun-teers 60% more likely (OR 1.63, CI 1.36–1.94) to receive treatmentthan those in villages using village government personnel, whenadjusted for other factors in the model and for the clustering of youngchild coverage within households (Table 3). More intensive contactwith the project team also affected treatment behavior, with house-holds randomly selected for interviews and eye exams in each villagemore likely to get their children treated. Children in larger families
Azithromycin treatment coverage in Tanzania 171
VG group CV group VG CV Total
Village 1 Village 2 Village 3 Village 4 Village 5 Village 6mean mean
Population 2,464 3,620 2,728 2,721 1,473 1,577 2,937 1,923 14,582# Eligible HH 384 560 401 440 223 243 448 302 2,251
(Total HH) (497) (731) (571) (563) (310) (304) (600) (392) (2,976)# Ten-cell units 32 33 26 36 18 15 30 23 160# Neighborhoods 4 5 6 9 8 4 5 7 36
(Balozi/Mtaa) (8.0) (6.6) (4.3) (4.0) (2.3) (4.0) (6.0) (3.3) (4.4)Avg # pre-school 2.16 2.18 2.06 1.90 1.83 2.06 2.14 1.92 2.05
children/eligible HH
% HH with 40.7% 64.8% 31.6% 47.8% 30.7% 32.9% 47.1% 38.6% 43.6%children 7–15, (275) (439) (321) (342) (189) (189) (345) (240) (1,755)none in school
Oph
thal
mic
Epi
dem
iol D
ownl
oade
d fr
om in
form
ahea
lthca
re.c
om b
y U
nive
rsity
of
Con
nect
icut
on
10/2
9/14
For
pers
onal
use
onl
y.
table 2. Percentage of individualstreated by risk group.*
were more likely to receive treatment, as were younger children. Othervariables including gender of the child, and whether or not school agechildren were sent to school, were not predictors of treatment.
The second significant risk group is females (primary caretakers ofyoung children). Females are also the family members who commonlybring children for treatment. Predictors of coverage of women over 7 years of age were similar to those of pre-school children (Table 3).However, each of the associations was weaker.
Coverage of males over 15 was much lower, as previously shown inTable 2, and none of the factors modeled were predictive of treatment(data not shown).
Discussion Higher coverage rates for groups at highest risk of trachoma (young children and women) in trachoma-endemic villageswere achieved using self-selected community volunteers compared tovillage government personnel. Although both methods could be con-sidered community-based, self-selected community volunteers weresignificantly more effective in treating children in rural villages thanwere local government personnel. We postulate that the self-selectedcommunity volunteers may have been more highly motivated, and thusmay have provided a higher quality of interpersonal contact. Anecdo-tal evidence suggested that the volunteers were more likely to answerquestions and provide more complete explanations than village gov-
172 M. Lynch et al.
VG Villages CV Villages
n % Range (%) n % Range (%)
Pre-school children 2875 63.1 55.3–69.5 1744 73.2 67.4–79.2School-age females 710 56.2 48.4–59.1 486 61.5 53.1–78.6School-age males 677 45.2 40.0–50.2 496 52.8 41.6–64.1Women (15+ years) 1861 52.2 46.1–57.3 1248 61.9 56.3–68.2Men (15+ years) 1479 29.6 25.3–33.4 996 30.2 23.5–37.1Total Eligibles Treated 7602 51.6 4970 58.6 p = 0.02
*(Eligible hh only).
table 3. Predictors of treatmentcoverage for children aged <8 years(adjusted for household levelclustering of treatment).
Variable Children <8 years Females >7 yearsOdds Ratio (95% CI) Odds Ratio (95% CI)
Odds Ratio (95% CI)CV intervention (vs VG 1.63 (1.36, 1.94) 1.45 (1.25, 1.70)
group)Per additional pre-school child 1.17 (1.07, 1.28) 1.12 (1.05, 1.20)
in householdIn questionnaire sample (vs 1.57 (1.26, 1.96) 1.21 (1.00, 1.46)
not selected for sample)Child age (per additional year) 0.96 (0.94, 0.99) 1.02 (1.01, 1.02)
Oph
thal
mic
Epi
dem
iol D
ownl
oade
d fr
om in
form
ahea
lthca
re.c
om b
y U
nive
rsity
of
Con
nect
icut
on
10/2
9/14
For
pers
onal
use
onl
y.
ernment personnel. The fuller explanation may have motivated agreater degree of compliance. This theory is supported by the highercompliance among those households who were interviewed by theproject team for the sub-study, as they also provided another oppor-tunity for households to ask questions and get explanations.
A number of socio-political factors may make Tanzania a unique case for this comparison. The history of community participation incommunal activities under the socialist governments of the 1970’s and 80’s extended government influence into villages using balozileaders as points of contact. The balozi system is also used to collectlocal taxes, which may have negatively affected households’ percep-tions of the balozi leaders. These factors may limit the generalizabilityof our results, but it seems reasonable to assume that many householdsin many countries would make distinctions between local govern-ment “officials” and volunteering neighbors in community treatmentsettings.
The community volunteer recruitment process appears to haveresulted in much higher proportions of women workers in the CV vil-lages, and a proportionally larger workforce available for householdrecruitment. In addition, the requirement that community volunteersbring a cup of maize meal to the planning seminar as a symbolicdemonstration of commitment is likely to have contributed to select-ing for higher motivation. These factors argue that the time and effortinvolved in recruiting volunteers results in higher coverage.
The number of children in the household was an important predic-tor for treatment coverage. Increasing numbers of pre-school childrenin the household significantly increased the chances of at least one childin the household being treated. This may reflect an appreciation byparents of the increased risk of trachoma with more young children inthe household.
Younger children were also more likely to receive treatment, eitherbecause of a heightened awareness of trachoma in this age group byparents/caretakers, or perhaps due to the relative ease of transportingyounger children to the neighborhood treatment post. Older childrenhad other responsibilities in the fields, with cattle, or with school, andmay not have been as available. We had postulated that families whoreport sending school-age children to school may have higher treat-ment coverage, but this was not a factor. This may suggest that this per-sonalized approach reached across the community, and not just to thosewho seem to be more active in seeking education for their child.
Recruitment was poor among men, regardless of recruitment strat-egy used. This has been a consistent finding in treatment programs.2,12
In part, the men are less available for treatment as they are doing fieldor other work. Anecdotally, men are conservative about taking newtreatments when they are essentially asymptomatic.
Pre-school children are the main reservoir of trachoma in this popu-lation,13 and females are also at higher risk of trachoma disease, prob-ably due to their close contact with younger children.14 These two riskgroups thus form the key groups for trachoma control. Men generallyhave lower prevalence rates of active trachoma.10,15 This treatmentprogram targeted recruitment of entire households with young chil-
Azithromycin treatment coverage in Tanzania 173
Oph
thal
mic
Epi
dem
iol D
ownl
oade
d fr
om in
form
ahea
lthca
re.c
om b
y U
nive
rsity
of
Con
nect
icut
on
10/2
9/14
For
pers
onal
use
onl
y.
References1 Reed M, Blumer J. Azithromycin: A
critical review of the first azilideantibiotic and its role in pediatricpractice. Pediatr Infect Dis J. 1997;16:1069–1083.
2 Schachter J, West S, Mabey D, et al.Azithromycin in control oftrachoma. Lancet. 1999;354(9179):630–635.
3 World Health Organization. Controlof Major Blinding Diseases andDisorders (2). Fact Sheet No. 214.Geneva: WHO, 1999.
4 Dawson C, Jones BR, Tarizzo ML.Guide to trachoma control. Geneva:World Health Organization,1981.
5 West S, Taylor H. Community-basedintervention programs for trachomacontrol. Int Ophthalmol. 1988;12:19–23.
6 World Health Organization. Vision2020: The Right to Sight. Geneva:WHO, 1999.
7 World Health Organization.Community Directed Treatment withIvermectin: Report of a Multi-Country Study. Geneva: WHO,1996.
8 Kipp W, Burnham G, Bamuhiiga J,Weis P, Buttner DW. Ivermectindistribution using communityvolunteers in Kabarole district,Uganda. Health Policy Planning.1998;13(2):167–173.
174 M. Lynch et al.
dren, but while men’s participation is desirable, it is less likely to beepidemiologically essential for community trachoma control.
There are some limitations to our study. CV villages were purposivelyselected for smaller populations and this could have affected theresults. However, other factors were more likely to be important thansheer population size in determining recruitment outcomes. The VGvillage with the highest coverage was the largest village of all. Numberof eligible households per ten-cell unit, a likely measure of the burdenof recruitment for village government personnel, was not associatedwith household coverage. Neighborhoods in the CV villages were nomore compact than the VG villages, with many households scattered inthe periphery in both sets. Therefore the differences in population sizeseem unlikely to significantly affect the results.
We also investigated the effect of the selection of the household forthe questionnaire survey on treatment coverage. Not surprisingly, thisadditional contact with the project team, who informed the householdof the upcoming treatment, answered questions, and paid intensiveattention to the children’s eyes, did have an effect on treatment cover-age. Households selected for the survey were 56% more likely to haveall children treated than households not sampled. The logistic regres-sion model therefore adjusted for sample selection to prevent con-founding the results.
Overall, the coverage rates were disappointing, with less than 75% ofeligibles treated. In part, this was due to very low turnout by males, butthe schedule of only one day per neighborhood (a realistic schedule fornational programs) may also have reduced coverage.Allowing a longertime for distributing drugs in the village, for example by using com-munity health workers as is currently being done by the Tanzanianational program, also seems to help increase recruitment.
This study suggests that the additional effort recruiting volunteersfrom the community for motivating households to participate in treat-ment is likely to result in higher coverage, particularly among key high-risk groups for trachoma.
Oph
thal
mic
Epi
dem
iol D
ownl
oade
d fr
om in
form
ahea
lthca
re.c
om b
y U
nive
rsity
of
Con
nect
icut
on
10/2
9/14
For
pers
onal
use
onl
y.
Azithromycin treatment coverage in Tanzania 175
9 Oladepo O, Okunade A, BriegerWR, Oshiname FO, Ajuwon AJ.Outcome of two patient educationmethods on recruitment andcompliance with Ivermectin in thetreatment of onchocerciasis. PatientEduc Counsel. 1996;29:237–245.
10 West S, Muñoz B, Turner V, MmbagaB, Taylor H. The epidemiology oftrachoma in central Tanzania. Int JEpidemiol. 1991;20(4):1088–1092.
11 Thylefors B, Dawson CR, Jones BR,West S, Taylor HR. A simple systemfor the assessment of trachoma andits complications. Bull WHO. 1987;65:477–483.
12 Bowman RJ, Sillah A, Van Dehn C,et al. Operational comparison ofsingle-dose azithromycin and topical
tetracycline for trachoma. InvestOphthalmol Vis Sci. 2000;41(13):4074–4079.
13 Taylor H, Rapoza P, West S, et al.The epidemiology of infection intrachoma. Invest Ophthalmol Vis Sci.1989;30:1823–1833.
14 Congdon N, West S, Vitale S, KatalaS, Mmbaga BBO. Exposure tochildren and risk of active trachomain Tanzanian women. Am JEpidemiol. 1993;137(3):366–372.
15 Mabey DCW, Bailey RL, Ward ME,Whittle HC. A longitudinal study oftrachoma in a Gambian village:Implications concerning thepathogenesis of chlamydialinfection. Epidemiol Infect. 1992;108:343–351.
Oph
thal
mic
Epi
dem
iol D
ownl
oade
d fr
om in
form
ahea
lthca
re.c
om b
y U
nive
rsity
of
Con
nect
icut
on
10/2
9/14
For
pers
onal
use
onl
y.
Oph
thal
mic
Epi
dem
iol D
ownl
oade
d fr
om in
form
ahea
lthca
re.c
om b
y U
nive
rsity
of
Con
nect
icut
on
10/2
9/14
For
pers
onal
use
onl
y.