Ghana ANC Fertility Poster

KEYWORDS: Antenatal Care · Fertility Rate · Spatial Autocorrelation · LISA · Random Forest · Care Efficiency Index · Ghana · Ecological Study · STROBE

Background

Ghana achieved remarkable progress in skilled antenatal care (ANC) coverage over 34 years, yet national-level statistics mask persistent subnational fertility and care efficiency inequities, particularly in the Northern Belt.

This study is the second in a longitudinal series examining spatial determinants of maternal health service utilisation. Part I characterised district-level insurance non-enrolment clustering; Part II extends the temporal framework to ANC–fertility dynamics across all 16 regions.

Objectives

Characterise temporal trends and spatial clustering in ANC coverage and TFR across Ghana 1988–2022
Identify the critical TFR threshold associated with ANC coverage decline
Develop and validate a novel Care Efficiency Index (CEI) for subnational performance monitoring
Classify regions into actionable risk strata for policy targeting

Data & Setting

Source: Ghana Demographic and Health Survey (DHS) subnational estimates — 9 waves: 1988, 1993, 1998, 2003, 2008, 2014, 2016, 2019, 2022.

Unit of analysis: Region-year observations (n = 90; 14–16 regions per wave)

Administrative context: Ghana expanded from 10 to 16 regions (2018–2019); historical data propagated using boundary dissolution rules.

Reporting: STROBE guidelines for ecological studies.

Methods

Gini Coefficient
Inequality decomposition

Global Moran's I
KNN k=4, 999 perms

LISA Clusters
Rook contiguity

Random Forest
n=200, max_depth=6

Partial Dependence
Friedman (2001)

Z-score Quadrant
Risk stratification

CEI = ANC%/TFR
Novel composite

GeoPandas + Plotly
Choropleth maps

National ANC ↑ 83.1% → 97.7%; Northern Belt shows greatest absolute gain (+43 pp)

Key Findings

−87.5%

Decline in inter-regional Gini coefficient (0.070 → 0.009)
indicating near-complete ANC coverage equity by 2022

TFR 5.90

Critical threshold (Random Forest partial dependence)
above which predicted ANC coverage drops sharply to 85–90%

Moran's I = 0.570

TFR spatial clustering in 2022 (p<0.001) while ANC clustering
simultaneously dissipated (I=−0.107, p=0.496) — significant decoupling

2.2×

Care Efficiency Index gap: Greater Accra (CEI=30.8) vs. North East (CEI=14.3)
persists despite ANC coverage convergence

Spatial Clustering (LISA, 2022)

Rook contiguity weights, p<0.10 threshold, 999 permutations

ML Model Performance

Metric	Random Forest	Decision Tree
Train R²	0.960	1.000
Test R²	0.381	−0.296†
CV R² (5-fold)	0.203	—
RMSE (%)	5.22	—
Top predictor	Survey year 43.7%	TFR 63.1%

†DT Test R²<0: predicts worse than mean — overfitting benchmark only. RF: n=200 estimators, max_depth=6, 80:20 train/test split.

Partial Dependence (TFR Threshold)

Critical TFR threshold = 5.90; ANC falls sharply above this value. Northern Belt regions most affected pre-2008.

Care Efficiency Index (CEI)

The CEI = Skilled ANC (%) ÷ TFR provides a novel composite indicator benchmarking how efficiently a health system converts fertility burden into skilled antenatal contact.

Rank	Region	ANC%	TFR	CEI	Tier
1	Greater Accra	95.6	3.10	30.8	High
2	Eastern	94.4	3.49	27.1	High
3	Central	89.3	3.80	23.5	High
...	...	...	...	...	...
14	Savannah	63.7	4.36	14.6	Low
15	Upper West	68.9	4.67	14.8	Low
16	North East	59.2	4.14	14.3	Low

Pooled averages 1988–2022. High ≥21, Moderate 16–21, Low <16.

Risk Zone Stratification

Risk Zone	N Obs.	Mean ANC%	Mean TFR
Critical (Low ANC / High TFR)	55	59.6	5.78
Emerging (Low ANC / Low TFR)	8	74.3	3.12
Workhorse (High ANC / High TFR)	13	93.1	4.41
Resilient (High ANC / Low TFR)	14	94.8	3.06

55 of 90 (61%) observations are Critical, predominantly Northern Belt pre-2008.

Discussion Highlights

ANC–TFR decoupling: Supply-side coverage improvements driven by NHIS, fee-exemptions, and CHW scale-up have outpaced demand-side fertility transitions — a finding with direct policy relevance for the Northern Belt.
Spatial persistence: Despite near-universal ANC coverage in 2022, TFR clustering in the Northern Belt intensified (Moran's I = 0.570). Fertility inequity is spatially concentrated even as ANC inequity is not.
Critical threshold utility: The TFR 5.90 threshold offers a measurable policy target for fertility transition programmes in Northern Belt regions currently above this value (Savannah: 5.8, Northern: 5.4).
CEI as monitoring tool: The Care Efficiency Index enables cross-regional comparison adjusting for fertility burden — proposed for integration into GHS annual district health reviews and NHIA performance framework.

Policy Recommendations

Targeted fertility transition and female educational empowerment in Northern Belt regions (TFR > 5.0)
Adopt CEI as a complementary indicator in Ghana Health Service performance monitoring
Apply spatial risk zone classification for differential resource allocation across regions
Integrate spatial machine learning into annual district health data review cycles

Limitations

Ecological design — no individual-level causal inference
Small ML sample (n=90): moderate RF R²=0.381
Within-region heterogeneity suppressed in DHS subnational estimates
CEI requires validation against individual-level routine data
Administrative boundary changes introduce estimation uncertainty pre-2018

Conclusions

Ghana's 34-year ANC convergence is a landmark public health achievement. However, persistent TFR spatial clustering and the 2.2-fold CEI gap demonstrate that coverage equity ≠ care efficiency equity. Fertility-sensitive policy instruments, composite performance indicators, and spatial targeting are essential to translate coverage gains into equitable reproductive health outcomes.

Interactive Dashboard: Full 12-panel interactive analysis with choropleth maps, LISA clusters, ML outputs, and temporal animations available at the accompanying repository.
Data: Ghana DHS 1988–2022 (dhsprogram.com) | Code: github.com/[author]/ghana-anc-spatial
Contact: valentineghanem@gmail.com