We showed that after 3 repeated mass azithromycin treatments, the factors most strongly predictive of ocular chlamydial RNA were ocular discharge, missing the previous mass azithromycin treatment, having a sibling infected with ocular chlamydia, and living in a larger community. Even after 3 mass treatments, the clinical signs of trachoma were strongly associated with chlamydial infection. These findings were confirmed with analyses using ocular chlamydial DNA as the outcome, and were robust in models that accounted for community and household clustering.
Few studies have assessed for risk factors of ocular chlamydial infection before or after mass azithromycin treatments. Studies conducted before mass treatments have not shown consistent associations, though various studies have suggested the importance of unclean faces, age, household fly density, infected siblings, and absence of a latrine 
. The few studies conducted after mass azithromycin treatments have generally found that missing a previous mass azithromycin treatment and younger age are associated with chlamydial infection, though individual studies have also implicated the absence of a latrine, travel, and the number of infected and untreated children per household 
. Previous studies have assessed for DNA evidence of ocular chlamydial infection. In contrast, our primary outcome was ocular chlamydial RNA—the most sensitive test for chlamydia currently available 
The multivariate analysis revealed several risk factors for chlamydial RNA after repeated mass treatments. The association between missing the previous mass azithromycin treatment and ocular chlamydial infection confirms the importance of the Antibiotic component of the SAFE strategy. The association between ocular discharge and chlamydial infection seems to indicate that unclean faces are an important risk factor for infection even after mass treatments 
. However, it is also possible that ocular discharge is not on the causative pathway for ocular chlamydial infection, but is simply a result of being infected. The association between above-median community population and ocular chlamydia could indicate that chlamydial transmission is more likely in crowded communities, or could simply reflect the difficulty for chlamydial infection to fade away in larger communities 
. This result should be interpreted with caution, since the population variable was based on only 12 villages. It is important to note that although we used multivariate analyses, the observational study design cannot rule out the possibility that the observed associations are the result of unmeasured confounding.
Consistent with previous reports, we were unable to show that accessibility to water or latrine status were associated with either ocular chlamydia or clinical trachoma after mass antibiotics 
. Although a previous study did find that lack of a latrine was associated with ocular chlamydial infection after mass azithromycin treatment, that study found that the relationship became weaker over time, and by 12 months after the mass treatment (analogous to this study), was no longer significant 
. It is possible that our inability to detect an association with the E components of the SAFE strategy in the current study is a function of an insufficient sample size, or misclassification during the survey, or an insufficient follow-up period. Nonetheless, across several studies, missing mass antibiotic treatments does seem to be the most important predictor of chlamydial infection and trachoma, at least in the short term. Further investigation into the role of missed azithromycin treatments, including ways to improve coverage of mass treatments, is warranted.
Our results provide evidence that after repeated mass azithromycin treatments, trachoma programs may be able to target antibiotic treatments to those most likely to be infected. For example, TF and TI were independently associated with ocular chlamydial RNA after 3 mass treatments—not surprising, since infection causes the clinical signs of trachoma. Moreover, having an infected sibling was significantly associated with ocular chlamydial infection. This is consistent with previous reports that have shown that ocular chlamydial infection clusters by household 
. Taken together, these results suggest that treating the households of children with clinically active trachoma, as the WHO has suggested in the past, may be a reasonable way to target those individuals most likely to be infected after repeated rounds of mass azithromycin have already been distributed 
. In this study, treatment targeted to households in which any child was observed to have clinically active trachoma would have resulted in antibiotic distributions to only 55% (200/364) of households, but still would have covered 80% (28/35) of households with ocular chlamydia. However, the feasibility and cost-effectiveness of such a strategy remains to be determined, since targeting households with clinically active trachoma would require examinations of all children in the community. A simpler strategy might be to continue mass azithromycin treatments in areas with highly prevalent trachoma. In fact, the WHO now recommends 5 rounds of mass treatment for areas with hyperendemic trachoma, a strategy that would likely benefit the communities in this study 
As a secondary outcome, we assessed risk factors for having a positive chlamydial DNA test. These results confirmed the risk factors found for the primary RNA outcome, and also suggested that travel outside the community, or hosting a visitor from outside the community, may be associated with ocular chlamydial infection. Travel to and from outside communities could re-introduce ocular chlamydia into a treated household. Other studies have similarly found that travel could be an important risk factor for trachoma 
. Trachoma programs with knowledge of major migration episodes may choose to wait until after the travel has occurred before scheduling a mass distribution of azithromycin, or treat a large enough area in order to mitigate this potential problem. In addition, given that most travel was done over short periods of time, trachoma programs may increase coverage to most travelers simply by returning to households with absent members in several days time.
We also assessed for risk factors of the clinical signs of trachoma, TF and TI, as a secondary outcome. Most previous studies that have assessed for risk factors for clinically active trachoma have been conducted before mass azithromycin treatments have been initiated. In these studies, several risk factors have consistently been associated with clinically active trachoma, including indicators of poor face hygiene such as nasal discharge and the presence of flies 
, household fly density 
, distance to water 
, absence of a latrine 
, and number of children per household 
. Fewer studies have reported risk factors for clinically active trachoma after mass antibiotic treatments. These studies have generally demonstrated that younger age, lack of latrines, unclean faces, and missed mass azithromycin treatments are associated with clinically active trachoma 
. In this study, conducted after 3 annual mass azithromycin distributions, clinically active trachoma was associated with younger age and ocular discharge, two commonly reported risk factors. In addition, active trachoma was associated with having a visitor in the past month, and male gender. The significance of the association between TF/TI and male gender is unclear, as ocular chlamydial infection was not more common among boys. Past studies have more frequently demonstrated an association between clinically active trachoma and female gender 
, though several studies have noted more clinically active trachoma among boys 
. The generalizability of this finding is uncertain, and may simply reflect the specific sample of individuals in this study.
In both univariate and multivariate analyses, the results for the primary RNA outcome were almost always confirmed by the results for the secondary DNA outcome, but often differed from the results of the TF/TI outcome. This likely occurred because the RNA-based and DNA-based tests are related tests, in that both are testing for genetic evidence of chlamydia. The clinical examination, on the other hand, is a test for conjunctival inflammation, and is often discrepant with laboratory tests 
. In fact, roughly 43% of children had TF/TI in this study, whereas only 7% had RNA evidence of chlamydial infection, consistent with other reports that have shown that the clinical signs of trachoma persist for many months after chlamydial infection has been cleared 
. Nonetheless, there were 2 factors that were associated with all 3 outcomes: ocular discharge and missing the previous mass azithromycin treatment. The role of ocular discharge as a risk factor is unclear, since it may simply be a result of infection. However, that missing a previous mass azithromycin treatment was associated with all 3 outcomes suggests that this is an important trachoma risk factor, and suggests trachoma programs should make efforts to enhance antibiotic coverage.
This study has several limitations. We chose a random sample of children from each community, and therefore did not assess the clinical or infectious status of all children in surveyed households. This sampling strategy limited our sample size, which did not allow us to assess for the presence of weaker associations. Due to the observational study design, we are unable to comment on antecedent-consequent relationships. Other limitations of surveys apply, such as the potential for recall bias and misclassification errors.
In conclusion, we showed that after 3 annual mass azithromycin treatments in a region of Ethiopia with highly prevalent trachoma, ocular chlamydial RNA was associated with missing the previous mass antibiotic treatment, ocular discharge, larger community size, having an infected sibling, and the clinical signs of trachoma (TF and/or TI). These findings suggest that (1) maximizing antibiotic coverage and promoting face washing are important goals for trachoma programs; (2) larger communities may require more prolonged treatment compared to smaller communities; and (3) after repeated mass azithromycin treatments, trachoma programs could consider antibiotic distribution strategies that target children with clinically active trachoma and their siblings. Further research into the factors associated with chlamydial infection after repeated mass azithromycin distributions will be helpful to guide trachoma program activities after mass azithromycin distributions have begun.