A total of 6503 married/union women met the inclusion criteria for this study. The mean (standard deviation (SD) age in years of the women was 35.27(7.1). Out of the 6503, 1877(28.9%) had no unmet need for family planning, 2918(44.9%) had unmet need for limiting while 1708(26.3%) had unmet need for spacing. The mean (SD) age in years of the respondents with no unmet need was 33.13(6.7) while that for those with unmet need for limiting and spacing were 39.19(5.7) and 30.94(6.1) respectively. Figures 1, 2, 3, 4, depict percentages of the four main variables (religion, region, age category and educational level) that were of primary interest in this study and grouped according to the outcome variable (unmet need for family planning) and further grouped according to place of residence (urban and rural). Majority (40.1%) of the respondents who had no unmet need were of the Islamic religion for urban setting. Those with unmet need for limiting and spacing were of the Charismatic religious belief for both the urban and rural residence. Respondents without any religious background had the least type of unmet need across both residential types.
With respect to regional distribution, majority of those with no unmet need for family planning were from the Northern region of Ghana for both urban (24.9%) and rural (40.1%) areas. Similarly, women in the Northern region had the highest unmet need for spacing for both urban (17.9%) and rural (28.3%) areas. Concerning unmet need for limiting, the Greater Accra region reported the highest (18.8%) for the rural areas while the Eastern region had the highest (15.0%) for the urban areas.
There was a high cluster effect at the multivariable analyses level. There were 8.88 and 3.95 for limiting and spacing respectively, Table 1. The adjusted relative risk ratio results presented in Table 1, constitute one out of the four Models specified in the model building subsection, though it contains all the variables in the other sub-models. The final Model was arrived at after calculating the goodness of fit of all the Models using the likelihood ratio test statistic and the Akaike information criteria, as presented in Table S1 (Additional file 1). The best model was selected on the basis that it had the lowest value of the Akaike information criteria (AIC). As stipulated in (Additional file 1: Table S1, the more the significant variables were added to a Model, the better its fit. The Akaike information criteria was 8783.67 for Model-4 with its closest value being 8936.00 for Model-3, indicating that Model-4 is a better fit compared to Model-3, The difference between the two Models was 129.33. The likelihood ratio test for Model 4 compared to Model-3 was 164.33 with a p-value < 0.001, reinforcing the point that Model-4 is a better fit Model. Model-4 was therefore used for the final analysis. The calculated unobserved effect for the best fit Model (Model 4) was 8.88 implying a standard deviation of 2.98 for limiting. That for spacing was 3.95 implying a standard deviation of 1.99. The covariance between limiting and spacing was 2.73, an indication of a weak correlation (0.46) between them. Thus a 1-standard deviation of the random effects amounts to an exp. (2.98) = 19.69 and exp.(1.99) = 7.32 significant change in the relative risk ratio for limiting and spacing. Due to the type of model specified for these analyses, results are reported as relative risk ratios instead of odds ratios as expected if binary logistic regression is used for the analysis.
Demographic determinants of unmet need for family planning
Table 1, shows both the bivariate and multivariable multinomial mixed effects logistic regression analyses results. Under the socio-demographic grouping approach with the adjusted relative risk ratio, there was a reduced risk of unmet need for limiting against no unmet need. A reduced risk of 99.9% (RR of 0.01 (95% CI, 0.00, 0.03, p-value < 0.001) was observed for people from the Upper West region compared to that of the Volta region. Also, a reduced risk of 88.8% (RR of 0.12 (95% CI, 0.22, 0.60, p-value = 0.01) for respondents from the Eastern region was observed against respondents from the Volta region. Similar observations for unmet need for spacing were made except that the highest relative risk among the ten regions was the Ashanti and Eastern regions for limiting as compared to no unmet need. Though the relative risk for the Upper West region was 0.20(95% CI, 0.05, 0.84, p-value = 0.028) and that of the Eastern region and Ashanti regions were 0.25 (95% CI, 0.07, 0.94, p-value = 0.040) and 0.80 (95% CI, 0.21, 3.12, p-value = 0.750) respectively, only Upper West was significant. This implies that the risk for unmet need for women from the Volta region in relation to those from the Upper West and Eastern regions for limiting were 205 and 9 times higher. That of spacing were 5 and 4 times higher. In terms of the age category, those within 15–19 were used as the reference group. The observations made were that a change in age from lower to higher corresponds to an increase risk of an unmet need for limiting compared to no unmet need. For instance, the risk of women aged 20–24 years had a risk ratio of 1.73 (95% CI, 0.22, 13.34, p-value = 0.600). The risk ratio for age group 35–39 years was 133 (95% CI, 18.12, 977.18, p-value < 0.001). The opposite was the case for all the year groups compared to the 15–19 years respondents for spacing. The risk of respondents aged 20–24 compared to 15–19 when evaluated under unmet need for spacing gave an RR of 0.29 (95% CI, 0.12, 0.71, p-value < 0.007) and that of 35–39 had an RR of 0.17 (95% CI, 0.07, 0.40, p-value < 0.001) times the risk for no unmet need. This shows that respondents within 15–19 years group were 3 and 6 times more likely to develop the need for spacing as against the 20–24 and 35–39 age groups. All the other age groupings were similarly related.
Figure 5, contains the adjusted predictive probabilities of the types of unmet need for family planning according to regions and age categories of respondents. The marginal predictive probabilities for the unmet need for limiting is highest among respondents from the Volta region (0.80) followed by the Eastern region (0.66) with the smallest being the three regions of the Northern part of the country. For the age category, the marginal probabilities for limiting increased upwardly with a higher age. The predictive probability for wanting to limit was 0.86 for the 45–49 year group and as low as 0.02 among the 15–19 year group.
Socio-economic determinants of unmet need for family planning
Four socio-economic factors were identified to be statistically significantly associated with unmet need for family planning. These were respondent’s educational level, wealth index of household, respondent’s occupation and partner’s occupation. With respect to wealth index, only the richest and poorer respondents showed a significant difference. The middle and the richer were insignificant statistically when compared to the poorest with regards to unmet need for limiting. Respondents who were poorest have 2 times (95% CI, 1.36, 2.97, p-value < 0.001) the risk of having an unmet need for limiting compared to the poorer respondents. Under spacing, the richest had approximately 32% more risk. The poorer had 89%more risk than the poorest respondents. Educational level did not demonstrate any statistical significant difference for spacing. For limiting, respondents with primary and secondary education had about 2 times the risk with a p-value < 0.001. With regards to respondent’s occupation, those in the services and professionals had an RR of 16.92 times (95% CI, 4.78, 59.97, p-value < 0.001) and 6.06 times (95% CI, 2.05, 17.92, p-value = 0.001) risk of having unmet need for limiting compared with those without any work. The marginal probabilities for educational levels, presented in Fig. 5, shows that, the predictive probability for respondents classified under primary was the highest (0.47), followed by secondary (0.44) for limiting.
Socio-cultural determinants of unmet need for family planning
Respondents, religion and ethnicity were the only socio-cultural variables statistically significantly associated with unmet need for family planning. Religion was re-categorized into no religion, Orthodox, Charismatic, Islamic and Traditional for further analysis. The results showed that respondents without any religious affiliation had more than twice (with a p-value = 0.002) and 22% (with a p-value = 0.430) the risk of experiencing unmet need for limiting and spacing respectively compared to those with in traditional religion. Unmet need for family planning for the Charismatic group was approximately 3 with a p-value < 0.001 and 1.65 with a p-value < 0.011 times the risk for limiting and spacing than it was for traditional religion. From Fig. 2, a higher predictive probability (0.46) was observed for respondents without any religious affiliation and those with the Charismatic faith (0.46) for limiting.
Psychosocial and other determinants of unmet need for family planning
All the variables identified under this category were insignificant except those who reported infrequent sex, partner’s opposition to use of contraceptives and respondents fear of side effects. Respondents who fear contraceptive side effects were 3 times (95% CI, 2.28, 3.80 p-value < 0.001) at risk of having unmet need for limiting and 2.58 times (95% CI, 2.05, 3.24, p-value < 0.001) more likely to experience an unmet need for spacing when compared to respondents who do not fear contraceptive side effects. Respondents who had infrequent sex were 4.6 times more likely to want to limit and 2.4 times more likely to space their children than those who had frequent sex.