Prevalence, progress, and social inequalities of home deliveries in Ghana from 2006 to 2018: insights from the multiple indicator cluster surveys

Background: Delivery in unsafe and unsupervised conditions is common in developing countries including Ghana. Over the years, the Government of Ghana has attempted to improve maternal and child healthcare services including the reduction of home deliveries through programs such as fee waiver for delivery in 2003, abolishment of delivery care cost in 2005, and the introduction of the National Health Insurance Scheme in 2005. Though these efforts have yielded some results, home delivery is still an issue of great concern in Ghana. Therefore, the aim of the present study was to identify the risk factors that are consistently associated with home deliveries in Ghana between 2006 and 2017–18. Methods: The study relied on datasets from three waves (2006, 2011, and 2017–18) of the Ghana Multiple Indicator Cluster surveys (GMICS). Summary statistics were used to describe the sample. The survey design of the GMICS was accounted for using the ‘svyset’ command in STATA-14 before the association tests. Robust Poisson regression was used to estimate the relationship between sociodemographic factors and home deliveries in Ghana in both bivariate and multivariable models. Results: The proportion of women who give birth at home during the period under consideration has decreased. The proportion of home deliveries has reduced from 50.56% in 2006 to 21.37% in 2017–18. In the multivariable model, women who had less than eight antenatal care visits, as well as those who dwelt in households with decreasing wealth, rural areas of residence, were consistently at risk of delivering in the home throughout the three data waves. Residing in the Upper East region was associated with a lower likelihood of delivering at home. Conclusion: Policies should target the at-risk-women to achieve complete reduction in home deliveries. Access to facility-based deliveries should be expanded to ensure that the expansion measures are pro-poor, pro-rural, and pro-uneducated. Innovative measures such as mobile antenatal care programs should be organized in every community in the population segments that were consistently choosing home deliveries over facility-based deliveries.

The current study used datasets collected in three waves by the Ghana Multiple Indicator Cluster Survey (GMICS) in 2006, 2011, and 2017/2018. The GMICS is a cross-sectional survey conducted by Ghana Statistical Service (GSS) in collaboration with the Ghana Health Service (GHS), Ministry of Health (MOH), and the Ministry of Education [27]. The survey received funding and technical support from the United Nations International Children’s Emergency Fund (UNICEF) and other international donors [27]. The primary goal of the MICS surveys is to analyze key indicators that assist countries to produce data for use in national development plans, policies, and programmes. On top of that, the GMICS is intended to assess progress towards SDGs and other agreements signed internationally [27]. MICS surveys use a multi-stage stratified cluster design to select a probability sample of households and women (15–49 years). This approach was used to nationally survey women in urban and rural areas from the erstwhile ten administrative regions in Ghana namely, Western, Central, Greater Accra, Volta, Eastern, Ashanti, Brong Ahafo, Northern, Upper East, and Upper West. At the first stage based on the 2010 Population and Housing Census (PHC) of Ghana, enumeration areas (EAs) were randomly selected, becoming the primary sampling units (PSUs). Every household within the EA is listed to create a sampling frame and a sample of households was chosen in the second stage using systematic random sampling. Then reproductive-aged women were recruited from these selected households. A total of 7,795 women within the ages of 15 to 49 years who had delivered two years before the data collection for all the three waves participated in this study. The outcome variable is the place of delivery, therefore home delivery is the focal point for the present study. This variable was derived from the survey question asking the participants about the place of their child delivery two years before the start of the survey. Participants were specifically asked this question, “Where did you give birth to (name of child)?” The response format to this question were these: Home (“respondent’s home” and “other’s home”); Public medical sector (“Government hospital”, “Government clinic/health centre”, “Government health post” and “Other public”); Private medical sector (“Private hospital”, “Private Clinic”, “Private maternity home” and “Other private medical”); and Other. We assigned a value of “1” to the home response and all other options were assigned “0”. The explanatory variables in the models were selected after a review of the literature and their availability in the dataset [28–30]. The authors explored the following variables: age of woman, education, polygyny, wanted last-child, parity, antenatal care (ANC) attendance, previous child loss experience, health insurance, household wealth index, urban–rural residence, and region of residence. The ANC variable was recoded as 0–3 times (less than 3), 4–7 times, and 8 times and above. It would have been helpful to compare women who did not attend ANC at all with the other categories, but data on ANC attendance in 2006 revealed that only one woman indicated she did not have an ANC visit. Therefore, to make ANC effect on Home delivery comparable over time, we decided to group those with no ANC attendance with those who had 1 up to 3 visits. We included the variable on respondent’s previous child loss experience in our models to ascertain its association with giving birth to their children in the home. It is not clear from the dataset or the questionnaire whether the experience of child loss occurred in a health facility or the home or any other place. All these variables were available in all three datasets except that of health insurance which was available in 2011 and 2017/2018; we included this variable because of its policy implication on maternal and child health. We did not include in our model the variable on religious affiliations of the respondents because it had no data on it in the most recent GMICS dataset (the 2017/18 dataset). As indicated in Table ​Table1,1, participants responded to all the variables using simple response options. Summary statistics of sociodemographic correlates and home deliveries in Ghana, 2006 to 2017–18 The datasets were cleaned, and variables recoded in STATA version 14. We accounted for survey weights for the differential probability selection of the sample. The variances were calculated to adjust for clustering, stratification, and design effects using the Taylor linearization technique [31]. We first conducted specific survey waves (2006, 2011, and 2017–18) univariates analyses, computing frequencies and percentages of all variables (Table ​(Table1—second,1—second, fifth, and eighth columns). Secondly, bivariate analyses were performed with Chi-square test of independence, estimating the relationship between the explanatory variables and outcome variable (place of delivery – home or facility delivery) as presented in Table ​Table1.1. Lastly, multivariate analyses with robust Poisson regression models incorporating all explanatory variables were used to model the prevalence of home delivery as well as examine its relationship, regardless of statistical significance in the bivariate analyses as presented in Table ​Table2.2. Because Poisson regression is applied to a binary variable, the robust error variance technique is used to avoid overestimating the error of the estimated prevalence ratio (PR). The preference for prevalence ratio over odds ratio is adequately explained elsewhere [32, 33], and the same thing applies to our study. The prevalence ratio and the adjusted prevalence ratio are reported. Sociodemographic correlates regressed on home deliveries in Ghana, 2006 to 2017–2018 0.995 [0.843,1.174] 1.242* [1.038,1.486] 0.810* [0.660,0.994] 1.333* [1.054,1.685] 0.985 [0.774,1.253] 1.036 [0.774,1.386] 0.924 [0.796,1.072] 1.100 [0.976,1.238] 0.751*** [0.633,0.890] 1.040 [0.893,1.211] 0.951 [0.764,1.183] 1.112 [0.900,1.374] 4.802*** [2.644,8.723] 1.677 [0.993,2.832] 12.11*** [6.998,20.95] 2.131** [1.216,3.736] 5.724*** [3.465,9.457] 1.786* [1.070,2.982] 3.706*** [2.032,6.759] 1.462 [0.868,2.463] 7.018*** [4.006,12.30] 1.671 [0.960,2.908] 4.724*** [2.852,7.824] 1.782* [1.061,2.994] 2.498** [1.379,4.525] 1.251 [0.759,2.061] 4.284*** [2.411,7.609] 1.461 [0.830,2.573] 3.317*** [2.075,5.303] 1.705* [1.039,2.800] 1.193 [0.937,1.520] 1.119 [0.908,1.379] 1.122 [0.841,1.497] 0.876 [0.686,1.117] 0.837 [0.645,1.086] 0.889 [0.670,1.180] 1.574*** [1.209,2.049] 1.201 [0.942,1.531] 2.025*** [1.488,2.756] 0.943 [0.715,1.244] 1.413* [1.032,1.934] 1.051 [0.758,1.457] 1.099 [0.966,1.252] 1.042 [0.932,1.166] 1.049 [0.904,1.218] 1.119 [0.975,1.285] 1.109 [0.935,1.316] 1.045 [0.886,1.232] 1.190 [0.960,1.475] 1.337** [1.103,1.621] 2.056*** [1.526,2.771] 1.832*** [1.384,2.426] 1.193 [0.900,1.582] 1.242 [0.941,1.641] 1.537*** [1.279,1.848] 1.398** [1.139,1.717] 2.945*** [2.263,3.833] 2.263*** [1.684,3.042] 1.604*** [1.239,2.078] 1.231 [0.907,1.670] 2.829*** [2.226,3.594] 1.605*** [1.322,1.950] 4.349*** [3.472,5.447] 1.767*** [1.412,2.211] 4.626*** [3.480,6.149] 2.443*** [1.808,3.301] 1.864*** [1.466,2.370] 1.291** [1.077,1.547] 2.167*** [1.753,2.679] 1.294* [1.057,1.583] 1.857*** [1.418,2.430] 1.302* [1.001,1.692] 1.217*** [1.091,1.359] 0.936 [0.841,1.040] 1.513*** [1.302,1.758] 0.957 [0.830,1.104] 1.340* [1.055,1.702] 1.049 [0.864,1.272] 1.678*** [1.430,1.968] 1.161* [1.017,1.325] 1.981*** [1.643,2.390] 1.517*** [1.283,1.794] 7.974*** [4.873,13.05] 3.441*** [1.967,6.016] 24.26*** [10.24,57.46] 6.689*** [2.376,18.83] 11.12*** [6.076,20.34] 4.240*** [2.248,7.999] 6.997*** [4.240,11.55] 3.254*** [1.858,5.700] 16.41*** [6.826,39.44] 5.703*** [2.060,15.79] 8.768*** [4.810,15.98] 3.617*** [1.950,6.707] 5.606*** [3.329,9.439] 3.077*** [1.742,5.434] 11.67*** [4.789,28.46] 5.505** [1.995,15.19] 6.740*** [3.671,12.37] 3.222*** [1.697,6.118] 2.616*** [1.506,4.545] 1.908* [1.097,3.319] 5.747*** [2.273,14.53] 3.303* [1.217,8.962] 3.913*** [2.061,7.428] 2.509** [1.298,4.850] 2.935*** [2.252,3.826] 1.504** [1.174,1.928] 3.796*** [2.939,4.903] 1.846*** [1.428,2.387] 3.033*** [2.137,4.305] 1.670** [1.205,2.314] 3.717*** [2.162,6.388] 1.491 [0.984,2.261] 3.401*** [1.803,6.414] 0.940 [0.619,1.428] 3.094** [1.482,6.460] 1.683 [0.756,3.750] 3.236*** [1.873,5.590] 1.336 [0.811,2.201] 3.339*** [1.794,6.217] 1.186 [0.821,1.715] 3.717*** [1.790,7.717] 2.040 [0.922,4.515] 3.462*** [2.016,5.947] 1.198 [0.781,1.839] 3.290*** [1.693,6.392] 0.958 [0.609,1.508] 4.625*** [2.277,9.393] 1.699 [0.778,3.711] 3.587*** [2.118,6.076] 1.307 [0.856,1.995] 1.976 [0.977,3.999] 0.836 [0.517,1.352] 3.157** [1.528,6.523] 1.489 [0.667,3.322] 2.475** [1.401,4.373] 1.244 [0.803,1.926] 2.235* [1.175,4.250] 0.851 [0.553,1.311] 2.691* [1.249,5.801] 1.712 [0.750,3.905] 2.624** [1.439,4.783] 1.088 [0.681,1.736] 3.498*** [1.831,6.683] 0.908 [0.616,1.337] 1.968 [0.867,4.464] 0.984 [0.419,2.309] 4.010*** [2.316,6.945] 1.232 [0.795,1.908] 5.674*** [3.109,10.36] 1.094 [0.751,1.593] 6.135*** [3.065,12.28] 2.087 [0.948,4.597] 3.459*** [1.973,6.066] 1.143 [0.731,1.787] 3.096*** [1.650,5.807] 0.633* [0.419,0.958] 0.747 [0.311,1.792] 0.329* [0.125,0.865] 4.350*** [2.568,7.371] 1.273 [0.823,1.970] 3.494*** [1.872,6.522] 0.746 [0.498,1.118] 2.831** [1.326,6.044] 1.022 [0.440,2.373] Exponentiated coefficients; 95% confidence intervals in brackets *p < 0.05, ** p < 0.01, *** p < 0.001 where p1 is the prevalence of delivery at home. We repeated these processes for all the three survey waves used in this study. Statistical significance is determined using 95% confidence intervals (CIs) and an alpha value of 0.05. The study was performed in accordance with the Declaration of Helsinki and approved by appropriate ethics committee. Trained field enumerators collected data on behalf of UNICEF and GSS. The MICS team of UNICEF-Ghana, the Ethical Review Board of the Ghana Health Service, and the Ghana Statistical Service approved the study. Informed consent was obtained from all respondents, and assent was obtained for respondents younger than eighteen from parents/guardians/adult household member before data collection. More details regarding the data and ethical standards are available at: https://mics.unicef.org/surveys. Therefore, ethics approval for this study was not required since the data is secondary and is available in the public domain.

N/A

• Dima Health