i Trends and contributing factors of the change of HIV prevalence over time among reproductive age group women in Ethiopia evidenced by Ethiopian demographic and health survey (EDHS) data: a cross-sectional study: Multivariate decomposition analysis.

s Background: Human immunodeficiency virus remains the leading cause of morbidity and mortality throughout the world. At the beginning of the epidemic, around 76.1 million people were infected and 32 million people died from AIDS-related illnesses in the world. Sub-Saharan Africa regions are the most affected regions and accounted for 67% of HIV infections worldwide, and 72% of the world’s AIDS-related deaths. Objective: To show trends and contributing factors for the change of HIV prevalence over time among reproductive age group women in Ethiopia. Methods: This study was conducted based on Ethiopian Demographic and Health Surveys data. A total of 10423 in 2005, 15153 in 2011, and 14159 in 2016 women were involved in the study. Multivariate decomposition analysis was performed using the mvdcmp Stata package to identify the contributing factors of change of HIV prevalence over time. The 95% confidence interval was used for the test of significance. Results: This study showed that 90.4 % of the change in HIV prevalence over time was attributable to behavioral change over time, particularly in women who were rural residents and not exposed to media. The behavioral change of women who live in rural areas was the major factor for the decline of HIV for the last ten years. The behavioral change of women who hadn’t exposure to media contributed 98.4% to the decline of HIV prevalence over the past ten years. Conclusion: The prevalence of HIV among reproductive age group women in Ethiopia was significantly declined over the last ten years and the decline was due to behavioral change over time. The major factor for the reduction of HIV prevalence overtime was the behavioral change of rural resident women. Therefore Ethiopian government should primarily focus on the strengthening and scaling up of behavioral change packages related to HIV prevention and control methods.

to provide comprehensive, universal, equitable, and affordable health services for the rural population. The program was provided as sixteen packages focused on health promotion and education supported by demonstration targeting households, particularly mothers and women through the house to house visits (6). based on promotive, preventive, and basic curative services for both communicable and non-communicable diseases (6). sampling was described in the complete EDHS report (9)(10)(11).

Variables
The outcome variable of this study was having HIV in their blood (Yes/No). Whereas place of residence, region, marital status, educational level, occupation, sex of household head, desire for children, heard about HIV, and media exposure were considered as independent variables.

Statistical analysis
After the data set was downloaded from the Measure DHS website (http://www.dhs program.com) the variables of this study were extracted using STATA version 14.2. Before any statistical analysis, the data were weighted using sampling weight for probability sampling and non-response to restore the representativeness of the survey and get reliable statistical estimates.
Trends and decomposition analysis STATA version 14.2 was used for Editing, recoding, descriptive and multivariate decomposition analysis. After the data were cleaned, categorized, coded, and weighted, we explored the descriptive statistics by using the frequencies and percentages and presented them by using tables and graphs. The trend of HIV prevalence among reproductive age group women was shown by using different characteristics. The trend period was divided into three phases; the first phase

Y = F(Xβ) = logit(Y) = Xβ
Where represent the dependent variable represents a set of predictor variables denote set of regression coefficients The proportion difference in Y between the two surveys of A and B can be decomposed as

Let the recent 2016 EDHS and reference 2005 EDHS datasets can be denoted by A and B
respectively.
For logistic regression, the log-odds or logit of the prevalence of HIV is given by (12).
represents endowments, which is explained by characteristics. An endowment is a change in HIV prevalence due to differences in characteristics. denotes coefficients or effect of characteristics which is unexplained (12). The coefficient is the change in HIV prevalence due to the effect of predictor variables.
The equation can be presented as: Where

Characteristics of the Study Population
Based on socio-demographic reports of EDHS data, more than 75% of women were living in rural   In the third phase (overall phase) it was increased by 2.3%, 1.9 %, and 1.1% point change in Addis Ababa, Dire-Dawa, and Afar regional state respectively. On the other hand, the overall change in a decrement of HIV prevalence based on the region was higher in Benishangul-Gumuz regional state with 1.8% point change followed by southern nation nationalities and people of Ethiopia (SNNP) regional state with 1.2% point change (Table 2).

Decomposition analysis
Overall from 2005 to 2016, there has been a significant decline in the prevalence of HIV in Ethiopia. The overall decomposition result showed that 90.4% of the decline in the prevalence of HIV over time was due to a difference in the effects of characteristics (behavioral difference or change) between the surveys. About 9.6% of the decline was due to differences in characteristics (compositional factors) but the change due to differences in characteristics (compositional factors or change in population proportion) was not significant (Table3).
Residence of women and exposure to media showed a significant effect on the decline of HIV prevalence over time. Keeping compositional changes constant, changes in the behavior of women who live in rural areas contributed more than 100% to the decline of HIV prevalence over the past ten years as compared to women who live in the urban areas. Similarly, behavioral change of women who were exposed to media contributed 98.4% to the decline of HIV prevalence for the last ten years as compared to women who were not exposed to media (Table 3).

Discussion
Human immunodeficiency virus infection remains the leading cause of morbidity and mortality throughout the world. At beginning of the epidemic, more than 76.1 million people were infected and 32 million people died from AIDS-related illnesses in the world (2). Ethiopia is one of the Sub-Saharan countries and is known for the epidemic of HIV.
The overall prevalence of HIV among reproductive age group women in Ethiopia was found 1.45,  (14). The country also has committed to reducing new HIV infections by 50 percent by 2020 and to ending AIDS as a public health threat by 2030. This is reflected in the Country's Health Sector Transformation Plan II 2015-2020, where one of the major indicators is a reduction of HIV incidence rate (15).
When we decompose the change of HIV prevalence over time, the behavioral change of the respondents between the surveys contributed 90.4% to the decline of HIV prevalence over the last ten years. Surprisingly behavioral change of rural resident women contributed more than 100% to the change of HIV prevalence over time as compared to urban residents. This could be due to the increment of access to health facilities, access to transportation, and the government's commitment to improving awareness of the community through health education and enabling them to use health services. In addition to this increased access to information and education over time through health extension workers help rural residents to know about the negative health consequences of HIV (16,17).
The behavioral change of women who were not exposed to media contributed 98.4% for the change of HIV prevalence over the last ten years. Even though they didn't have access either to Television, Radio, Newspapers, or Internet, now a day's information is near to their house in all rural areas of Ethiopia. Because, in addition to health extension workers, many governmental and non-governmental organizations work closely with the community to reduce the prevalence and negative impacts of HIV by giving health education about HIV prevention, and control methods (18,19).

Strength and limitations of the study
This study had several strengths. The first is, the study was based on nationally representative large datasets and all estimates of the study were done after the data were weighted for probability sampling and non-response. Therefore it had adequate statistical power and generalized to all women in Ethiopia. Lastly, multivariate decomposition was applied to identify factors contributing to the change of HIV prevalence over time in Ethiopia. As a limitation, some variables were not collected in all EDHS's data like wealth index, and are not used for decomposition analysis.

Conclusion and recommendations
The prevalence of HIV among reproductive age group women in Ethiopia was significantly declined over the last ten years and the decline was due to differences in coefficients (behavioral changes) between the surveys. Behavioral changes of women who live in rural areas and are not exposed to media over time were the source of the decrease in HIV prevalence over the last ten years in Ethiopia. Therefore Ethiopian government should primarily focus on the strengthening and scaling up of behavioral change packages related to HIV prevention and control methods.  (20).

Consent for Publication
Not applicable

Availability of Data and Materials
The EDHS data sets are open and can be accessed from the Measure DHS website (http://www.dhs program.com) through an online request by explaining the objective of the study. The datasets analyzed during the current study are available from the corresponding author upon reasonable request.

Conflict of Interests
The authors declare that they have no competing interests.

Funding
We did not receive any funds for this research.

Authors' Contribution
YN, DA, MS, WW, AA, SG, BB, TY, and GF were involved in this study from the inception to design, acquisition of data, data cleaning, data analysis and interpretation, and drafting and revising of the manuscript. YN prepared the final draft of the manuscript. All authors read and approved the final manuscript.