Induced abortion among adolescent girls and young women: should geography matter in Ghana?

The study addresses rising abortion rates in certain regions, including sub-Saharan Africa, and focuses on Ghana where abortion laws are relatively liberal but social stigma persists. It explores how geography—operationalized as ecological zones (savannah, coastal, forest) and place of residence (urban/rural)—relates to induced abortion among adolescent girls and young women (AGYW) aged 15–24. Guided by an ecological framework that considers individual, community, and environmental influences on behavior, the study fills a gap by jointly examining ecological zone by place of residence among AGYW who have ever been pregnant. The hypotheses were: (a) AGYW in the Savannah zone are less likely to have an induced abortion than those in the forest zone; and (b) AGYW residing in urban areas are more likely to have an induced abortion than those in rural areas.

Prior research links abortion outcomes to legal contexts, with liberal laws associated with better health outcomes and restrictive laws with negative outcomes. In Ghana, despite a favorable policy environment for comprehensive abortion care, criminal connotations and stigma persist, affecting access and experiences. Geographic disparities influence sexual and reproductive health service utilization, with differences by ecological zones and rural/urban residence. Previous studies typically treated ecological zone and residence as separate predictors and primarily examined all women of reproductive age, despite evidence that induced and unsafe abortions are prevalent among women aged 15–24. Factors repeatedly associated with induced abortion include age, education, marital status, parity, contraceptive use, and household income.

Design and data source: Secondary analysis of the nationally representative 2017 Ghana Maternal Health Survey (GMHS), implemented by the Ghana Health Service and Ghana Statistical Service with ICF technical support. Ethical approval for the survey was granted by ICF and GHS ethics committees; data use permission was obtained from ICF. Study design was cross-sectional. Sample: Weighted sample of 3194 AGYW aged 15–24 who had ever been pregnant (as only they could have experienced live birth, miscarriage, or induced abortion). Outcome variable: Induced abortion within the five years preceding the survey, coded 1 if yes and 0 otherwise. Independent variables: Individual-level factors: age (15–19, 20–24), educational attainment (no education, primary, middle/JHS, secondary/higher), marital status (currently married, cohabiting, not in union), current contraceptive use (yes, no), age at first sex (<18, ≥18), parity (no birth, 1, 2+). Contextual-level factors: ecological zone (savannah, coastal, forest) and place of residence (rural, urban). Zones reflect Ghana’s 10-region classification at the time: savannah (Northern, Upper West, Upper East), coastal (Western, Central, Volta, Greater Accra), forest (Brong Ahafo, Ashanti, Eastern). Statistical analysis: Descriptive statistics (frequencies, percentages) characterized the sample; chi-square tests and bar charts depicted induced abortion across zones and residence. Multilevel modeling used two-level multilevel binary logistic regression (melogit), with clusters treated as random effects to account for community-level variability. Women were nested in households, and households within clusters; random intercept models were estimated. Four models were fitted: Model 0 (empty), Model I (individual-level variables), Model II (contextual-level variables), Model III (both levels). Adjusted odds ratios (aORs) with 95% CIs were reported. Model comparison used log-likelihood and AIC, with survey weights and svy commands applied to account for complex sampling and ensure generalizability. A separate set of models partitioned the data by ecological zone, controlling for individual-level factors, to estimate the effect of residence (urban vs rural) within each zone.

• Prevalence: Induced abortion among AGYW (15–24) was 26.0%.
• Descriptive patterns by geography: A higher proportion of induced abortions occurred among urban residents across all ecological zones. Urban forest zone had the highest proportion (38.0%); rural forest (23.1%), rural coastal (21.9%), urban savannah (12.0%), rural savannah (7.0%).
• Individual-level associations (Model III aOR [95% CI]):
  • Age 20–24 vs 15–19: 2.14 (1.57–2.92), higher odds.
  • Education: Secondary/Higher vs none: 2.38 (1.51–3.74); Primary and Middle/JHS showed elevated odds in simpler models but were attenuated in the full model.
  • Marital status: Cohabiting vs married: 2.32 (1.63–3.29); Not in union vs married: 2.04 (1.43–2.92).
  • Contraceptive use: No vs yes: 0.54 (0.43–0.67), lower odds.
  • Age at first sex: ≥18 vs <18: 0.28 (0.21–0.38), lower odds.
  • Parity: 1 vs none: 0.06 (0.04–0.08); 2+ vs none: 0.04 (0.03–0.06), lower odds.
• Contextual-level associations (Model III):
  • Urban vs rural: 1.83 (1.47–2.28), higher odds.
  • Ecological zone: Coastal vs savannah: 2.60 (1.89–3.62); Forest vs savannah: 2.74 (1.92–3.91), higher odds.
• Zone-specific residence effects (partitioned models, adjusted for individual factors):
  • Forest zone: Urban vs rural: 1.91 (1.23–2.94), statistically significant.
  • Coastal zone: 1.41 (0.97–2.02), not statistically significant.
  • Savannah zone: 0.90 (0.49–1.68), not statistically significant.
• Random effects and model fit: Intraclass correlation (ICC) decreased from 0.15 in the empty model to 0.07 in the full model, indicating remaining but reduced between-cluster variability; AIC and BIC decreased across models, favoring the full model.

Findings support both hypotheses: AGYW in the savannah zone were less likely to have induced abortions than those in the forest zone, and urban residents were more likely than rural residents. The particularly elevated risk among urban-dwelling AGYW in the forest zone suggests context-specific drivers such as migration from savannah areas to forest regions for economic activities (e.g., mining, head porter work) that may increase vulnerability to sexual exploitation and unintended pregnancy. Urban environments may also entail greater exposure to risky sexual behaviors and easier access to abortion services, alongside more autonomy and less adherence to traditional norms discouraging abortion. Comparisons with prior studies from Ghana and other African settings indicate similar urban–rural and zonal disparities, reinforcing the ecological framework wherein individual and community factors jointly shape abortion experiences. Additional significant correlates—older age within the 15–24 range, cohabiting or not being in union, higher education, later sexual debut, contraceptive use, and higher parity—align with plausible mechanisms related to knowledge, aspirations, access to resources, stigma avoidance, and reproductive intentions.

There is a strong association between ecological zone, place of residence, and induced abortion among AGYW in Ghana. Urban-dwelling AGYW in the forest ecological zone have the highest likelihood of reporting induced abortion. Interventions should prioritize AGYW in urban areas of the forest zone, providing enhanced health education and sensitization about induced abortion. Strengthening campaigns to increase contraceptive use among AGYW nationwide, particularly in the urban forest zone, is recommended. Future research should employ longitudinal designs and incorporate additional potential confounders (e.g., cultural norms, specific health variables) to better understand causal pathways and improve validity.

Cross-sectional design limits causal inference between residence/ecological zone and induced abortion. Reliance on self-reported data introduces potential social desirability and recall bias. Analysis was constrained to variables available in the GMHS; important factors such as cultural values/norms and health variables (e.g., gestational hypertension) were unavailable as potential confounders.