We have applied principal component analysis and factor analysis methods to analyze the perception of households towards a community-based insurance scheme (iCHF). Both methods reduce many variables (statements) into a fewer number of factors. PCA assumes there is no unique variance thus the total variance is equal to the common variance while FA assumes that the total variance can be partitioned into common and unique variances.
The results for the two methods differ somewhat for the number of factors identified and how much each factor explains of the total variance. However, the most important perception factors are the same across the two methods; Convenience (location and opening hours of iCHF offices, etc.), Quality (healthcare services), Preferences (the importance of alternative risk-reducing strategies such as saving and borrowing) and Knowledge.
Our findings partly contrast former studies on community-based insurance and household perception factors. Jehu-Appiah et al., (2012), in a study from Ghana, identified scheme factors (premiums, scheme benefits, and scheme convenience) as the most important perception factors [17]. In our study, the same factors, except for scheme convenience, were not important. Kansra and Gill (2017), in a study conducted in India, identified “lack of awareness and information about the insurance scheme” and “low and irregular income” as the most important perception factors [18]. In our study, however, the statements concerned with affordability (price-income statements) did not turn out as important. A possible explanation for this could be due to differences in study settings of the three studies. The study in Ghana was conducted in both rural and urban areas and the study in India was conducted in urban areas while this study was conducted in rural areas.
As concerning the multivariate regressions (logistic), we find that the quality of care, access to the iCHF offices, and preferences are the factors with the most significant influence on iCHF membership status. Furthermore, the presence or non-presence of household characteristics did not impact our results in important ways. The only socio-demographic variables that turned out significant, in combination with the perception factors, were age and income. Surprisingly, education did not turn out significant for any of the regressions performed. One possible explanation for this finding is because the education level of the respondent is not representative of the education level of the household (the average education level). Also, the variation in education was small across the respondents. Furthermore, for the regression that considers household characteristics alone, gender was significant (p = 0.03), however, when including the perception factors, gender became insignificant. This last finding may suggest confounding effects between the perception factors and gender.
Our findings concerning provider quality indicate that people are more willing to purchase insurance if the quality of health care services is improved. This finding is consistent with results from other research conducted in Tanzania. Several studies have identified a positive association between quality of care and the enrollment into the predecessor of the iCHF scheme [10, 15, 26]. Similar findings have also been reported for Uganda [27] and Kenya [28].
Another interesting finding is that the statements about the role of prices (premiums) and low income (affordability) were not important. This suggests that purchasing power is not an important barrier for enrolling in the iCHF in Tanzania. The answer to one of the statements, not included in our factor analysis, seems to confirm this. From the survey it follows that 93% of the respondents strongly agreed or agreed to the following statement; “the ICHF scheme will become more important to me if additional health care expenditures were covered despite a corresponding increase in the premium.” Furthermore, 2/3 of all respondents agreed or strongly agreed with the statement “the iCHF premiums are affordable to me.”
Access to the iCHF offices (location, opening hours, and modality of collecting membership card) is the most important scheme factor in our analysis. This finding is in line with Winani (2015) who found that a longer distance between the community and the nearest CHF office acted as a barrier to enroll in the health insurance scheme in Tanzania [29]. Other studies from Africa also confirm such effects[17, 30]. The factor concerned with beliefs and alternatives, confirms as expected that, respondents that consider alternatives to insurance (saving and borrowing) and cure (traditional healers, health is in the hands of God) are less likely to be members of iCHF. The sign of the factor that includes recommendations from relatives, friends, and iCHF representatives turned out opposite of what was expected. A possible explanation is that the recommendations given to the respondents from family and friends are not very enthusiastic, in this way affecting their enrolment decision negatively.
The results from the multivariate regressions performed by Jehu-Appiah et al. (2012) and Kansra and Gill (2017) confirm that the most important perception factors also became the most important determinants in the regression analyses [17, 18]. The study from Ghana found the benefits of the insurance scheme, the premiums, and convenience to be important while factors related to the quality of care were not associated with insurance scheme enrolment [17]. The study from India, on the other hand, identified a lack of awareness and low and irregular income as the most important determinants [18]. Thus, our findings differ from both studies since provider quality is important while affordability (income and premiums) is not important. As concerning household characteristics, our study identifies age and income to have some relevance, while in [17] most household characteristics (education, income, gender, age, and religion) became significant while[18] did not identify any household characteristics (gender, age, income, marital status, and education) as being significant. The two studies differ somewhat from our study since [17] surveys a mix of urban and rural populations with more than 60% of the respondents being males, while [18] surveys urban populations with 91% of the respondents being males. Our study, in contrast, study rural populations (mainly farming households) and 58% of the respondents were females.
Limitations and strengths
A cross-sectional study is without some limitations. This study was conducted in two districts of Tanzania within one region, which makes it difficult to generalize the interpretation of the results to the other regions implementing the iCHF scheme. We, therefore, argue that the findings should be interpreted with some caution. Furthermore, a majority of the respondents were female (58%) thus introducing the possibility for gender bias. We can not rule out that female respondents differ from male respondents along some dimensions. However, our survey had a participation rate equal to 100%, meaning that we are not confronted with any selection bias.