The evaluation of the December 2007 MDA in Orissa led to the description of several predictors of and barriers to compliance with the MDA program. These predictors and barriers were used to refine a pre-MDA community-based educational campaign that was then implemented in six blocks in Khurda District, three of which had received the early version of the campaign―the Com-MDA+LM area―and three of which were new to the campaign―the Com-MDA area. The results were remarkable. In the Com-MDA+LM areas MDA compliance increased from a 2007 baseline of 59.5% 
to 90.2%, well above that target of 80.0% compliance among the entire population. There was also a marked increase in compliance in the Com-MDA areas to 75.0%, which is close to the target and much improved from the 2007 baseline of 52.2% 
. It is likely that the baseline MDA compliance for rural areas in this district is around 52%, and it is clear that both intervention groups had a significant increase in adherence over that baseline.
This study not only makes an important and direct contribution to the effort to interrupt the transmission of LF in India, it also serves as an example that can be used by other programs to overcome barriers to MDA compliance in affected populations. The KAP survey allowed identification of predictors of and barriers to adherence to a DEC regimen Factors identified in the previous evaluation were targeted by an educational campaign delivered one month prior to the 2008 MDA. The increased adherence during the 2008 MDA campaign provided not only the proof-of-concept that the targeted educational program worked, but it also validated the previously identified predictors and barriers. This assessment demonstrates how critical operational research is to any health program, particularly one whose success depends on changing health behaviors. Fortunately, this research can lead to simple and effective solutions. Developing messages that address key concepts for improving compliance with the MDA program is essential. In Orissa, these include: 1) making people aware of the occurrence of the MDA in advance of its occurrence―the CASA program is launched one month prior to the MDA, 2) making people aware of the purpose of the MDA medication, 3) making people aware that everyone is at risk for infection, 4) making people aware that one can be infected and still feel well, and 5) making people aware that side effects of DEC are infrequent and mild. Additionally, data from those who did not take DEC suggested that the medication's benefit needs to be personalized. The person who takes the medication needs to feel that they or a close family member stands to benefit directly. Lofty national goals did not speak to those who did not take DEC in this evaluation population. Individualized programs will need to be developed to address the specific needs of each location.
One unique and important finding from the 2008 evaluation is that community-based lymphedema management programs positively affect MDA compliance independently of such programs' effects on the other predictors of compliance. Even after multivariable modeling controlling for all of the other LF and MDA knowledge predictors, knowing any one of the three components of the management of leg lymphedema predicted adherence to the DEC regimen. This positive impact of community-based lymphedema management education persisted even among those who had no household members with lymphedema. Additionally, the Com-MDA+LM area had the highest level of persons adhering to the DEC regimen in this study (90.2%). Admittedly, part of the explanation may be that the Com-MDA+LM had received a pre-MDA educational campaign two years in a row, but the campaign in the year 2007, which did not focus on predictors of adherence, was largely ineffectual (as evidenced by DEC adherence of 59.5% in the area in 2007). Previous authors have suggested that morbidity control programs could improve MDA compliance 
, but this study is the first to provide data wholly consistent with, if not unequivocally substantiating, that hypothesis. Perhaps these programs are effective because they help maintain awareness of LF and its chronic manifestations in the community and reinforce LF messages taught in the pre-MDA programs. Or perhaps they enhance trust at the community and individual level by providing programs benefiting a generally marginalized and stigmatized population, those who suffer from lymphedema and elephantiasis. Lymphedema management programs could provide an ideal platform for both LF and MDA education to improve MDA program compliance. As India approaches LF elimination, there will be a continued need to assist LF patients with clinical disease. Integrating lymphedema management with LF elimination efforts could be a more cost-effective way to ensure that MDA compliance remains high, even if political pressure to continue funding elimination efforts diminishes.
There are several factors that could influence compliance that merit further comment. Persons in the Com-MDA+LM areas had the fewest number of people who reported no education and the highest number who reported reading well. In univariable analysis reading well influenced the decision to take DEC and education level had relatively little impact on the decision; in multivariable analysis the relationship reversed. Possibly the ability to weigh the risks and benefits of MDA compliance is more directly related to education level than to literacy. Additionally, the mechanisms utilized to distribute the educational message included many verbal routes (i.e. street plays, auto-rickshaws, etc). However, an assessment of literacy, or health literacy, using a validated tool might allow a more thorough examination of this complex relationship. Multivariable analysis suggested that those with less education were more likely to comply. Perhaps those with less education are more likely to accept public health messages. It is important to note that although the relationship with education level is statistically significant, because of smaller numbers in each group the confidence intervals around the ORs are wide and in many cases approach one. It may be that the impact of education on compliance is much less than suggested by our analysis; this is an issue that should certainly be examined in future studies.
Knowing a household member with leg edema could also influence one's perception of risk for LF. While the prevalence of leg edema in a household member did not differ statistically across the three groups, the highest prevalence was reported in the Com-MDA+LM group. Whether this represents actual increased prevalence or increased recognition of the condition because of the lymphedema management program is unclear. Even though this factor was found to be a predictor of MDA compliance in univariable analysis, it was not significant in multivariable analysis. One possible reason for this is that the CASA educational message emphasized that everyone was at risk for infection and that one might be infected even if one felt well.
Finally, there was an interaction between knowing about the MDA in advance and knowing everyone was at risk for LF. Those who only knew everyone was at risk for LF had a small increase in MDA compliance. Those who only knew about the MDA in advance had a larger increase. Those who knew both had a synergistically larger increase. Why this was so is not clear. Perhaps those who understood both messages had a heightened sense of benefit or felt more empowered to achieve their own health goals because they felt at risk for infection and that they had the opportunity to avail themselves of preventive medication. Perhaps the interaction reflects the influence of another factor, such as an understanding of side effects and their management. In either case, the interaction points to the importance of addressing risk of LF and opportunity to access the beneficial MDA medication in any educational message.
The limitations of this evaluation are similar to other retrospective evaluations that use the EPI random walk cluster method. Selection bias was reduced by defining a strict set of rules governing household selection and replacement of non-participants was not allowed. The evaluation was cross-sectional, so causality cannot be assumed. However, given that most of the predictors identified in this evaluation were the same as in the prior evaluation and that the knowledge of the predictors in the Com-MDA+LM area was higher in this evaluation that in the prior one, it is likely that the predictors are causal. The generalizability of the results may be limited to rural areas as urban areas were not included. Finally, although there is a definitive baseline MDA compliance for the Com-MDA+LM area, the baseline for the Com-MDA and the MDA-only areas are based on less direct empirical data. The Com-MDA baseline is derived from the Bologarh MDA compliance of 52.2% for the 2007 MDA. The fact that the compliance in Banapur for the 2008 MDA was 52.9% suggests that MDA compliance in rural areas of Khurda District is approximately 50–55%.
Determining the predictors and barriers of adherence to the DEC regimen distributed in the MDA allowed for identification of key educational messages that were incorporated into a pre-MDA community-based LF educational campaign and resulted in a marked increase in regimen adherence. An added benefit was demonstrating that community-based lymphedema management programs independently enhanced adherence. Although further work is needed to determine exactly which components of lymphedema management programs influence MDA program compliance, one should not wait for those results before investing in such programs which address the twin goals of improving the lives of those suffering from filarial disease and increasing compliance with MDA programs to the level needed for the interruption of LF transmission.