Optimising care for uncomplicated type 2 diabetes mellitus in Lagos, Nigeria: cost and benefit estimates using real-world data

Abstract

Introduction

Diabetes mellitus accounts for a significant share of morbidity and mortality in ages 30–70 years worldwide. In sub-Saharan Africa, diabetes care is often suboptimal for reasons ranging from health system weaknesses to patient illiteracy and non-compliance with recommendations. This study explores the potential costs and health benefits of optimising care for uncomplicated type 2 diabetes in Lagos State, Nigeria.

Methods

Longitudinal data on medical care patterns and resource use (consultations, medications, diagnostics and lifestyle counselling) over a 1-year period were collected retrospectively from 84 health facilities in Lagos. Medical resource prices were obtained from a subsample of 26 facilities. Patient care gaps were assessed by comparing actual journeys to official diabetes management guidelines. Mixed-effect regression analyses were employed to explore the impact of care elements on blood glucose control and model the potential complications averted if all patients received recommended care, with extrapolation to the entire Lagos population.

Results

Data from 642 patients with uncomplicated type 2 diabetes were analysed. A one-unit increase in consultation score (a measure of the adequacy of consultation visits) and having health insurance coverage were linked to 47-unit and 29-unit lower blood glucose levels, respectively. Optimising diabetes care requires additional US$357 per patient annually, totalling US$206 million statewide, with medications comprising 73% of costs. Enhanced care could reduce stroke and myocardial infarction by 2% (12 675 cases) and 4% (22 282 cases) over 7 years, respectively, at a cost of US$5893 per complication averted.

Conclusions

There are significant potential health and economic benefits from the optimisation of diabetes care in the State, but realising these benefits will require substantial additional investment alongside efficiency gains in medicines procurement and care delivery. There is a need to explore innovative financing and delivery options, including digital value-based care interventions and cost-saving care approaches such as pooled medication procurement, while also investing in local medicines production capacity and expansion of the health insurance coverage.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Diabetes mellitus is a major driver of morbidity and mortality worldwide, accounting for 6.7 million deaths globally in 2021. In sub-Saharan Africa, diabetes care remains suboptimal due to health system weaknesses, limited patient adherence and financial constraints. Existing global evidence highlights the cost-effectiveness of improved diabetes care delivery models, but there is generally limited granular evidence on the economics of diabetes care in sub-Saharan African settings.

WHAT THIS STUDY ADDS

• Using real-world data, this study reveals significant disparities between actual patient care and the guideline recommended practices and underscores the significant role of financial protection and regular follow-up in achieving better blood glucose control. It quantifies the financial investments required to improve care delivery, showing drugs as a major contributor to the cost burden and therefore a prime target for cost reduction interventions. We also demonstrate that there are huge potential returns on such care optimisation investments in the forms of reductions in risks of diabetes-related complications, such as stroke and myocardial infarction.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The results highlight significant financial requirements for scaling up optimal diabetes care in Lagos, Nigeria, necessitating innovative financing mechanisms and cost-saving strategies. Policymakers should consider expanding health insurance coverage, implementing pooled medication procurement and strengthening local pharmaceutical manufacturing. Additionally, digital value-based care interventions could enhance care delivery efficiency, making diabetes management more accessible and sustainable in Lagos and similar urban settings.

Introduction

Diabetes mellitus (DM) is a chronic non-communicable disease (NCD) of endocrine and metabolic dysfunction that accounts for a significant share of morbidity and mortality in people between 30 and 70 years old, accounting for 6.7 million deaths globally in 2021.¹ In the same year, estimates of the International Diabetes Federation (IDF) showed that worldwide, 537 million adults had diabetes and this number was expected to increase to 783 million by 2045.¹ In low- and middle-income countries (LMICs) particularly, there has been a rapid increase in the prevalence of DM while it had been classified among the so-called ‘Diseases of the West’, due to factors such as increasing life expectancy, urbanisation and changes in lifestyles, especially diet.² In Nigeria, the prevalence of DM increased from 2.2% in 1992 to an estimated 5.8% in 2018, though this is reasonably an underestimation, as an estimated 66.7% of people with DM in Africa are thought to be undiagnosed.³ Most of the patients with DM in Lagos, Nigeria, are between the ages of 41 and 60.⁴

Chronic uncontrolled blood glucose (sugar) in people with DM is associated with several complications including cardiovascular disorders, kidney diseases, eye-related disorders, neuropathy, rheumatoid arthritis, osteopathologies and foot diseases, the risks of which are significantly reduced with optimum sugar control.^{5 6} This underscores the need for regular clinical contact for close monitoring of sugar control and disease progression. Apart from the global guideline (WHO), Nigeria applies national (Federal Ministry of Health of Nigeria’s) and subnational (eg, Lagos State’s) guidelines for clinical management of DM. Optimum diabetes management requires regular supply of medicines, diagnostic facilities, data collection tools, qualified healthcare workers, informed patients and services adapted to the cultural and social peculiarities of the population.⁷

In LMICs, patients with DM often do not receive the care they need to halt disease progression. From a supply side perspective, shortage of trained healthcare personnel, poor infrastructure, weak supply chains and poor quality of services, if even available, are among the factors that render health systems too weak to handle the rising prevalence of NCDs.^{7 8} From the demand side, such factors include limited access to DM care and/or poor adherence to medication and lifestyle recommendations, which can be attributed to low patient education, cultural barriers and financial constraints.^{9 10} Moreover, traditional or spiritual healers and informal medicine vendors may be the first resort for people seeking care and evidence shows such treatments impact negatively on adherence to orthodox medications and recommendations.^{10 11}

DM care imposes a huge economic burden on households, communities and nations. Global diabetes-related health expenditure accounts for 11.5% of worldwide health spending.¹ Available evidence from sub-Saharan Africa indicates that annual national direct costs varied from 3.5 billion to 4.5 billion International dollars.¹² The IDF estimates suggest that Africa spends 7% of its healthcare budget on DM,³ and studies have attributed figures ranging from 14% to 90% of all treatment costs to drugs.¹³ Despite these costs, DM care has been shown to be cost-effective with such care elements as intensive glycaemic control, statin therapy, routine eye examination, smoking cessation, comprehensive foot care, patient education and diabetes self-management education shown to be very cost-effective or cost-saving.^{14 15} This provides economic impetus for strategic investments in diabetes care. Moreover, unmet diabetes care need has been linked with substantial productivity losses which directly undermine national macro-economic prosperity,¹⁶ and the economic impact of diabetes disproportionately affects people of low socioeconomic groups.^{17 18}

Locally grounded evidence is crucial to guide investments in DM care.⁸ In sub-Saharan Africa, evidence around the economics of diabetes care is particularly scarce, as a recent review of the literature in LMICs showed that the majority of the available studies were from Asia and Latin America.¹² An earlier Medline search of studies on the economics of diabetes in sub-Saharan Africa over a 20-year period had identified only one study, highlighting the dearth of evidence on DM economics in this setting and the need for more research work on this subject.¹⁹ Moreover, much of the available evidence is based on modelling studies.^{20 21} Compared with most of the existing studies that rely primarily on modelling assumptions, use of routine care data reflects better the realities of service delivery, patient behaviour, and provider cost structure, thus generating context-specific evidence that is more actionable for policymakers than abstract models alone. In this study, we contribute to the existing knowledge base by determining the costs and health effects of optimising care for uncomplicated type 2 diabetes care in Lagos State, Nigeria, using actual health facility patient records and on-the-ground costing data.

Methods

This study builds on the study by Banigbe and colleagues²² which evaluated the quality of care provided for patients with DM and hypertension in Lagos State, using retrospectively extracted longitudinal data of patients visits at 84 health facilities in Lagos, Nigeria over a 1-year period (January–December 2019). That study was performed as a baseline for a broader impact evaluation study to examine supply and demand side factors associated with the implementation of the Lagos State Health Scheme (LSHS) of the Lagos State Health Management Agency (LASHMA). For the current study we used the longitudinal data of DM patient visits and additionally collected costing data among a representative sample of health facilities.

Study context

Nigeria is a lower-middle-income nation with a gross domestic product per capita of 2162 USD, and a population of approximately 202 million. Lagos State, situated in the South-West, is the second most populous state in Nigeria. The capital of the State is Lagos, Africa’s largest city, with an estimated population of up to 21 million.²³ To enhance universal health coverage, Nigeria launched the National Health Insurance Scheme in 1999, introducing a policy of decentralisation to the states in 2015. Lagos established the LSHS under LASHMA, becoming one of Nigeria’s first State Health Insurance Schemes. Additionally, private Health Maintenance Organizations (HMOs), licensed by the National Health Insurance Authority (NHIA), provide healthcare services for a set fee. Despite these efforts, insurance coverage remains limited.²⁴

Study population

Patients with an established diagnosis of diabetes and who had at least two clinical visits between January and December 2019 were eligible to have their records abstracted. The exclusion criteria were: (1) Patients less than 30 years old (to exclude type 1 diabetes); (2) Patients with established cardiovascular disease or complications of diabetes, for example, heart failure, stroke, myocardial infarction (MI), diabetic nephropathy, at the first visit in 2019; (3) Pregnant women (to exclude gestational diabetes). The use of age 30 cut-off aligns with previous type 2 DM prevalence studies in sub-Saharan Africa which applied 20 years,²⁵ 25 years,^{26 27} 30 years,²⁸ 35 years²⁹ and 40 years.^{30 31} Moreover, in the AABBCC clinical tool for distinguishing diabetes types published in the 2024 American Diabetes Association Standards of Care in Diabetes, diagnosis at ‘age <35 years’ is considered a pointer to type 1 diabetes.³² We excluded patients with documented macrovascular or microvascular complications to ensure that the cost and benefit estimates reflected optimisation of uncomplicated type 2 diabetes management pathways.

Sampling approach

The sampling frame for Banigbe et al²² included all facilities eligible to participate in the LSHS. A two-stage stratified random sampling technique based on SPSS random sample generator was used to select the 84 study facilities, spanning LSHS-empaneled and non-LSHS-empaneled, public and privately owned, secondary and primary health facilities. In this study, we included all (43) facilities that provided care to patients with diabetes in 2019 out of the original 84 facilities. About two-thirds (60.5%) of the facilities were empaneled in the LSHS; secondary health facilities comprised 65.1% of the sample, while 55.8% were privately owned health facilities.²² The facilities cut across 19 (out of 20) local government areas of Lagos State.

Data collection

Patient-level clinical data were abstracted during the year 2019 from individual patient medical records (both paper and electronic) by trained research associates who were qualified medical doctors, using customised forms. Furthermore, a trained data collector visited selected health facilities and collected price data pertaining to health interventions and resources consumed in DM care journey from a subsample of 26 healthcare facilities between March and May 2023. We selected facilities to capture the heterogeneities across ownership type (public vs private), LASHMA empanelment (yes/no) and level of care (primary vs secondary).

Clinical data

Clinical data collected from patient records encompass, among others, patient demographics, dates of visits, consultations, history taken, physical examinations conducted, prescribed medications, counselling provided, ordered laboratory tests and corresponding laboratory test results for each visit the patient had. Based on structured and non-structured data abstracted from the records, we estimated for each patient the total yearly quantities (in milligrams or international units) of various drugs consumed, including insulins, glibenclamide, metformin, gliclazide, glimepiride, linagliptin, galvusmet, pioglitazone, vildagliptin and sitagliptin.

Cost data

Costs in this study are defined as the direct medical costs of DM care. Cost data collected pertain to unit prices of routine consultations, DM medications, laboratory tests (fasting blood glucose [FBG], urinalysis, glycated haemoglobin, and serum electrolyte/urea/creatinine) and lifestyle counselling. At each facility, prices collected for each item covered LSHS prices, HMO prices and out-of-pocket (OOP) prices. Our intent was to estimate the total financial requirements of delivering optimal care within the health system as currently structured, where a substantial share of services is financed directly by patients. Based on price information from the 26 healthcare facilities, average unit prices were computed per type of facility type/facility level/payment category (LSHS/HMO/OOP). As the drugs and dosages even for the same drugs were different for different patients, we standardised the prices by computing the per milligram or per international unit equivalents of the prices of the oral hypoglycaemics and insulins, respectively.

For each category with a missing item price, the average cost of the closest category was imputed. For instance, where the average OOP price for an item was not available for public primary facilities, the average OOP price at public secondary facilities was imputed. As costs of lifestyle counselling were generally stated as ‘free’ in public facilities, the cost of a consultation was imputed for cost of lifestyle counselling in these types of facilities. All costs were converted from 2023 Nigerian Naira to 2023 US Dollars (USD) using the conversion rate as published by Central Bank of Nigeria on the last business day of April 2023, which is halfway through the cost data collection period of March–May 2023.³³ The unit prices of all the resources included in costing are presented in online supplemental material 1.

Analyses

All analyses were conducted using Microsoft Excel 365 V.2305 and STATA V.16.1. Patient and facility characteristics were summarised and presented using descriptive statistics: nominal variables were described using absolute and relative frequencies, and continuous variables were described by means and standard deviations (SD).

Computing cost of optimising care

The cost requirement for optimising DM care was defined as the difference between the cost of the present observed provided care and the cost of providing the ideal (best practice) DM care for the patients. The best practice journey definition was based on the framework developed by Banigbe et al²² and validated by a team of experts through a Delphi process. This framework combined three contextually relevant guidelines: (1) The WHO HEARTS Technical Package308 was developed in 2016 for use in primary care settings of LMICs; (2) The Lagos State standard treatment guideline developed by the State Ministry of Health in 2018 specifically for managing patients enrolled in the LSHS; (3) The Clinical Practice Guidelines for Diabetes Management in Nigeria (second edition) co-developed in 2013 by the Diabetes Association of Nigeria and the Endocrine and Metabolism Society of Nigeria. Components of this model included in our cost analysis are presented in table 1. Items such as blood pressure check were considered as part of ‘consultation’ and were therefore not costed separately.

Table 1. Overview of an ideal outpatient care journey for a diabetes mellitus patient in a year.

Journey component	Minimum number per year
Consultations (routine follow-up visits)	12
Lifestyle counselling	1
Laboratory investigations
Fasting blood glucose	12
HbA1C	3
Fasting lipid profile	1
Electrolytes, urea and creatinine	1
Urinalysis for proteinuria	1
Medication (insulin or oral antihyperglycaemic) prescription	Per visit
Eye examination	1

Framework adapted from a combination of three contextually relevant guidelines: (1) The WHO HEARTS Technical Package308 developed in 2016 for use in primary care settings of low- and middle-income countries; (2) The Lagos State standard treatment guideline developed by the State Ministry of Health in 2018 specifically for managing patients enrolled in the LSHS and (3) The Clinical Practice Guidelines for Diabetes Management in Nigeria (second edition) co-developed in 2013 by the Diabetes Association of Nigeria and the Endocrine and Metabolism Society of Nigeria.

HbA1c, glycated haemoglobin; LSHS, Lagos State Health Scheme.

For each patient, the actual yearly total care cost was computed by first identifying and quantifying the resources used in the care year and multiplying each by the average unit prices of the components (subject to facility attended and insurance coverage status), in line with standard costing guidelines.³⁴ The daily amount of medications (in milligram for oral hypogylcaemics and international units for insulins) consumed by a patient was determined through the analysis of the patient’s prescription—the medications, the amount (strength) per dose and the dosing frequencies. In line with common practice in Nigeria, it was assumed that a patient receives a 1 month’s equivalent of medication (daily amount times 30 days) in each monthly hospital visit. The total annual amount consumed was then estimated by multiplying the monthly amounts by the number of monthly consultations recorded. Further, the cost of the ideal annual journey for the patient was estimated by multiplying the minimum volumes of resource as recommended in the guidelines by the unit prices (subject to facility attended and insurance coverage status). Again, the cost of medications was based on the patient’s own prescriptions, assuming 100% adherence by the patient. In other words, the ideal scenario assumes that a patient takes the medications prescribed for him or her consistently for 365 days of the year. The total annual amount of drugs was calculated by multiplying the daily requirements (already derived in the actual scenario estimation) by 365 days. Adjustments were made in the calculations where drugs were dropped, substituted or added anytime in the study period. Average actual journey cost, average ideal journey cost and average cost gap were computed for the full study population and stratified per facility type (public/private and primary/secondary) and health financing mechanism (HMO/LSHS/OOP).

Exploring the determinants of glucose control

To explore the determinants of the glucose control, we employed mixed effects (restricted maximum likelihood) regressions. Mixed-effects models were used to appropriately analyse repeated observations nested within individuals and health facilities. By incorporating both fixed and random effects, these models account for inter-patient and inter-facility variability, improve statistical efficiency and reduce bias compared with traditional regression approaches that assume independence of observations.³⁵ The model is specified in the following equation:

$${\displaystyle {\displaystyle FB{G}_{it}={\beta }_0+{\beta }_1{X}_{it}+u_i+{ϵ}_{it}}}$$

Where FBG_it is the FBG for individual i at time (or visit) t; is the intercept; Xit is a vector of explanatory variables for individual i at time t; is random effect for individual i; and is residual error term for individual i at time t.

A broad range of explanatory variables spanning demographic factors (gender and age),³⁶ health insurance coverage status,³⁷ use of insulin (yes/no),²⁰ comorbidities and measures of completeness of care journey (compared with the guidelines)³⁸ was considered for inclusion in the mixed-effects regressions. For each component of the care journey, a completeness score was computed as a continuous variable (eg the actual number of visits as a fraction of the ideal number of visits) or a binary variable (eg, having at least one fasting lipid profile test done (1) or not (0)). A series of univariate regression analyses was used to identify those that had significant relationships with the outcome variables for inclusion in the multivariate model. Beyond the univariate analyses, the selection of variables was also influenced by broader literature evidence, medical theory and clinical relevance. The development of the model followed a manual stepwise addition and deletion scheme,³⁹ guided by the Akaike information criterion values. Alpha was set at 5% level.

Estimating potential benefits of optimising DM care

Using the coefficients of the statistically significant journey completeness indicators from the regression, we modelled the blood glucose readings for each patient in a scenario of ideal patient journey. For the scenarios of actual journeys and ideal journeys, we computed the average (mean) blood glucose readings for each patient over the 1-year period. We then computed the proportions of patients (in both scenarios) that have their blood sugar controlled (defined as average FBG <126 mg/dL) and the proportions with uncontrolled sugar. Additionally, we estimated for both scenarios the incidence of strokes and MIs in the following 7 years in both scenarios. The difference in the incidences is the potential benefit from optimisation of DM care. Estimation of the risks was based on the findings of Boden-Albala et al, a prospective study of the relationship between FBG and the risks of ischaemic stroke and cardiovascular events in a multiethnic cohort of diabetics.³⁷ Boden-Albala et al followed, prospectively, cohorts of non-diabetics, DM patients with controlled FBG and DM patients with uncontrolled FBG over 7 years. They reported the subgroup sample sizes, global numbers of cardiac events and strokes and the risk ratios of same events in the different study groups. Combining the data provided, we worked backwards to derive the numbers of events and the risks (incidence proportion) of MI and strokes in the controlled and uncontrolled sugar groups. These risk estimates were applied to the corresponding groups in our study population—both in the documented actual journey scenario and the predicted (modelled) scenario of ideal patient journey. For each scenario, the total number of events is the sum of the numbers of events in the controlled and uncontrolled groups.

Extrapolating results to Lagos population

The average annual investment gap per patient and the risks of complications found in this study were extrapolated to the entire Lagos State level. We assumed a State total population of 21 million²³ with an estimated adult proportion of 50% and adult DM prevalence of 5.5%⁴⁰ and estimated a total adult DM population of 577 500 patients. By multiplying the total adult DM population by the mean annual per patient investment gap, we computed the annual statewide extra investment requirement. Lastly, we computed the cost per case of complication averted by dividing the total investment requirement by the total number of potential complications averted (ie, the difference in the potential numbers of adverse events between the actual scenario and ideal journey scenario).

Scenario analyses

Beyond the primary cost model which estimated the costs of achieving 100% compliance with the standard care guidelines, we modelled the costs and resource gaps for achieving (1) 75% and (2) 50% compliance with the care guidelines.

Patient and public involvement

Patients or the public were not involved in the design, conduct, or reporting, or dissemination plans of our research.

Results

Patient characteristics

Table 2 provides the descriptive statistics of the study patients and their care journeys. A total of 642 patients with type 2 DM were included in the analysis. The number of visits per patient ranged from 1 to 14, with a mean of 5.11 (SD: 3.30) visits while the mean follow-up period was 6.44 (SD: 3.43) months. The majority (71%) of the patients were between the ages of 35 and 60 years, and there were more women (57%) than men in the sample. About one-third of them received care in secondary facilities while one-third visited primary facilities. Among the 350 patients (55%) that had their FBG recorded at the first visit, the majority (75% [262/350]) had their blood glucose uncontrolled (table 2). Overall, the FBG records ranged from 46 mg/dL to 600 mg/dL, with a mean of 174.78 mg/dL (SD: 76.3). Out of a total of 1350 observations, 15 were above 400 mg/dL, including the single observation of 600 mg/dL which might represent a case of (impending) Hyperosmolar Hyperglycaemic State.

Table 2. Descriptive statistics of the patients and their care journeys.

Variable	Number of patients n (%)	Mean (SD)
Total	642 (100)	–
Gender		–
Male	274 (42.7)
Female	368 (57.3)
Age (years)		55.6 (11.6)
18–34	25 (3.9)
35–49	184 (28.7)
50–64	273 (42.5)
65+	160 (24.9)
Level of health facility visited		–
Primary	215 (33.5)
Secondary	427 (66.5)
Ownership of health facility visited		–
Public	281 (43.8)
Private	361 (56.2)
Insurance cover		–
Yes	184 (28.7)
No	458 (71.3)
Baseline (first visit) glucose status		–
Controlled (FBG <126 mg/dL)	86 (13.4)
Uncontrolled (FBG ≥126 mg/dL)	264 (41.1)
Undocumented	292 (45.5)
Insulin use		–
Yes	34 (5.3)
No	608 (94.7)
Number of hospital visits		5.1 (3.3)
Follow-up period (in months)	–	6.4 (3.43)
Consultation score	–	0.4 (0.3)
Actual FBG	–	174.7 (76.3)
Predicted FBG	–	123.6 (15.1)

FBG, fasting blood glucose; SD, standard deviation.

Patient journey costs and investment gap

Table 3 provides an overview of estimated average per patient cost of actual care journey, costs of ideal journey and the investment gap. The average cost of the actual patient journey is US$96 per patient per annum. In contrast, the average ideal journey cost is US$453 per patient per year. The average investment gap is thus US$357. The costs of medications are the largest driver accounting for 73% of total journey costs.

Table 3. Annual per patient costs of actual journey, ideal journey costs and investment gaps in 2023 USD.

	Total costs	Consultations	Medications	Eye exam	Diagnostics	Counselling
Actual journey	US$96	US$7	US$76	US$0.13	US$11	US$2
Ideal journey	US$453	US$23	US$331	US$17.32	US$73	US$9
Investment gap	US$357 (79%)	US$16 (70%)	US$255 (77%)	US$17.19 (99%)	US$62 (85%)	US$7 (78%)

Table 4 disaggregates the results of table 3 per facility ownership, facility level and health insurance status. In terms of facility ownership, the average actual cost and costs gap at the private facilities were US$131 and US$483 (respectively) compared with the average actual cost of US$51 and cost gap of US$195 for patients attending public hospitals. Patients visiting primary facilities incurred an average cost of US$54 with a cost gap of US$191 while those visiting secondary hospitals incurred an average cost of US$118 with a cost gap of US$441. Similarly, the average actual cost among the uninsured was US$87 with a cost gap of US$332 compared with the insured patients with an average actual cost of US$124 and a cost gap of US$386.

Table 4. Annual per patient costs of actual journey, ideal journey costs and investment gaps in 2023 USD, disaggregated by facility ownership, facility level and health insurance coverage status.

	Actual journey	Ideal journey	Investment gap	Actual journey	Ideal journey	Investment gap
	Private facilities			Public facilities
Consultations	US$12	US$38	US$26 (68%)	US$1	US$4	US$3 (75%)
Medications	US$103	US$447	US$343 (77%)	US$41	US$182	US$141 (77%)
Eye exam	US$0.23	US$31	US$31 (100%)	US$0	US$0	US$0 (0%)
Diagnostics	US$12	US$82	US$70 (85%)	US$9	US$60	US$51 (85%)
Counselling	US$3	US$16	US$13 (81%)	US$0.32	US$0.65	US$0.33 (51%)
Total costs	US$131	US$614	US$483 (79%)	US$51	US$246	US$195 (79%)
	Primary facilities			Secondary facilities
Consultations	US$1	US$4	US$3 (75%)	US$10	US$32	US$22 (69%)
Medications	US$45	US$179	US$134 (75%)	US$92	US$408	US$316 (77%)
Eye exam	US$0	US$0	US$0 (0%)	US$0.20	US$26	US$26 (99%)
Diagnostics	US$7	US$60	US$53 (88%)	US$13	US$79	US$66 (02%)
Counselling	US$0.33	US$1	US$0.47 (47%)	US$3	US$14	US$11 (79%)
Total costs	US$54	US$244	US$191 (78%)	US$118	US$559	US$441 (79%)
	Uninsured			Insured
Consultations	US$5	US$17	US$12 (71%)	US$14	US$37	US$24 (65%)
Medications	US$70	US$316	US$246 (78%)	US$96	US$340	US$244 (72%)
Eye exam	US$0	US$8	US$8 (100%)	US$0.45	US$41	US$41 (99%)
Diagnostics	US$11	US$72	US$61 (85%)	US$10	US$73	US$63 (86%)
Counselling	US$1.40	US$5	US$3.60 (72%)	US$3	US$19	US$16 (84%)
Total costs	US$87	US$418	US$332 (79%)	US$124	US$510	US$386 (76%)

Determinants of blood glucose control

The variables included in the final model and the mixed effects regression results are presented in table 5. Frequency of consultation (with prescription) and health insurance coverage were found to be significant predictors of blood glucose control. One unit increase in the consultation score (computed as the number of a patient’s consultation visits divided by 12) is significantly associated with an estimated 47 units (mg/dL) lower blood glucose level. Other significant associations were age and insulin use. Gender on the other hand showed no significant correlation with blood glucose levels; time from first visit was borderline significant. Diagnostics (singly or collectively) and comorbidity variables for instance were uncorrelated in univariate analyses and were not included in the multivariate model.

Table 5. Results of mixed effects regression model.

Number of observations	1350
Wald χ2 (6)	81.42
Prob>χ2	0.0000
Log restricted-likelihood	−7630.0249

FBG	Coef.	SE	z	p>l zl	95% CI
Time from first visit	−1.07	0.56	−1.89	0.059	−2.17	0.04
Age	−0.93	0.24	−3.90	0.000	−1.39	−0.46
Male gender	5.98	5.90	1.01	0.311	−5.58	17.54
Health insurance cover	−29.28	6.70	−4.37	0.000	−42.41	−16.15
Consultation score	−47.36	13.77	−3.44	0.001	−74.35	−20.37
Insulin use	40.40	9.82	4.11	0.000	21.14	59.65
Constant	252.27	14.44	17.47	0.000	223.97	280.57

AIC: 15261.04; BIC: 15313.12; intraclass correlation coefficient: 0.40.

AIC, Akaike information criterion; BIC, Bayesian information criterion; FBG, fasting blood glucose; SE, standard error.

Potential benefits of optimisation of blood glucose control

At the current care experience, 19% of the patients had an average FBG classified as controlled while 81% were uncontrolled over the study period. In a hypothetical scenario of full consultation (and prescription) and health insurance coverage for all the patients (modelled using the coefficients of the regression analysis), an estimated 51% would have controlled FBG while 49% would be uncontrolled. Based on Bolden-Albala et al,³⁷ we estimated 7-year stroke risks of 7% and 13% in patients with controlled and uncontrolled FBG, respectively, and 7-year MI risks of 16% and 28% in controlled and uncontrolled FBG, respectively. Applying these to our actual and hypothetical care scenarios, we estimated potential total stroke incidences of 78 (12%) and 64 (10%) in the actual and modelled scenarios, respectively (averted cases: 12 cases (2%)). Similarly, we found total potential MI incidences of 165 (26%) and 140 (22%) in the actual and modelled scenarios, respectively (averted cases: 25 (4%)).

Lagos population-level estimates

Extrapolating the findings to the entire Lagos population, we found a total annual investment gap in DM care of US$206 million. Optimising DM patient care in Lagos State would result in a potential 2% (12 675 cases) and 4% (22 282 cases) lower incidence of stroke and MI, respectively, over a 7-year period. These equate to US$5893 per averted case of complication.

Scenario analyses cost estimates

In a scenario of 75% compliance with the ideal stipulations of the guidelines, we estimate a mean annual journey cost of US$340, leaving an investment gap of US$244 (72%) compared with the estimated actual journey costs. Statewide, this corresponds to an annual investment gap of US$141 million. Similarly, to achieve a 50% compliance with the guidelines, our model estimates a mean annual journey cost of US$227, implying a mean deficit of US$131 (58%) and a population-level annual investment requirement of US$76 million. Further results from the scenario analyses are provided in online supplemental material 2.

Discussion

This study evaluated the cost and effectiveness of type 2 diabetes care in Lagos State, Nigeria. A retrospective analysis of patient records showed that current care costs US$96 per patient annually—79% lower than the US$453 needed for optimal care. Medications accounted for over 73% of total costs. The findings reveal significant gaps in care quality, limiting effective diabetes management. While best practice care could reduce diabetes-related strokes and MI at an incremental cost of US$5893 per averted case, it would require an additional US$206 million annually. This estimate represents the aggregate financial gap in direct medical spending needed to achieve guideline-concordant care across the entire system, combining current expenditures by government, insurers and households.

Our findings of an average actual cost of US$96 represents 5% of the annual per capita income which stood at US$1930 in 2023,⁴¹ and a state level extrapolated annual equivalent of US$56 million (if all DM patients accessed care). These figures are lower compared to previous Nigerian studies. For instance, based on a study conducted in 2012 at a tertiary facility in the Niger Delta region of Nigeria, Suleiman and Festus reported an average direct cost of illness per patient with type 2 diabetes US$284.57 per patient, amounting to an annual national direct cost of US$1.6 billion.⁴² Similarly, Abdulganiyu and Fola (2014) in their 2010 study at a tertiary institution in North-Eastern Nigeria estimated an annual average direct cost of illness of approximately US$320, representing 88% of the annual per capita income and an annual national burden of about US$1.5 billion.⁴³ Suleiman et al’s study conducted in 2004 at a tertiary facility in South-West Nigeria reported an average direct cost of illness of US$262.22 amounting to 84% of annual per capita income and an annual national direct cost of US$1.1 billion, based on a prevalence estimate of 3%.⁴⁴ Variations in the costs reported can be attributed to several factors, including inflationary influences associated with the temporal gap across the studies, methodological heterogeneities, variations in scope of cost elements included, differences in study settings (eg, tertiary facilities vs primary and secondary facilities), and case mix variations.^{13 45 46} Also, our population level cost burden is smaller than their estimates because we report single state-level cost burden while they report national-level extrapolations.

The finding of a cost of US$5893 per complication averted is substantial but this is in fact an overestimation, considering that the current study has only included MI and strokes. Left out in this analysis are numerous other diabetes complications avertable by optimum sugar control, including retinopathies, nephropathies, foot ulcers, and neuropathies. Studies have shown that excess costs of such diabetes complications are significant and the highest average cost ratios were recorded for nephropathy, diabetic foot and acute stroke while lowest cost ratios were recorded for retinopathy, ketoacidosis and hypertension.¹³ Further, mortalities averted were not considered, neither were productivity losses associated with diabetes morbidity and mortalities. Incorporating these elements in an appropriately designed economic evaluation model is likely to return economically more favourable results.

Furthermore, our primary model of cost gap assumes a scenario of 100% compliance with the guidelines for all patients with DM, including taking the prescribed medications for 365 days of the year, but this is rarely achieved in real-world settings. Evidence from a systematic literature review from high-income countries shows adherence to DM medications to be between 36% and 93%.⁴⁷ Poorer adherence is reported in LMICs due to a multiplicity of factors extending beyond cost of services⁴⁸ to include cost of reaching the health facility, cultural beliefs and lack of awareness.⁹ To reflect these realities, we conducted scenario analyses at 75% and 50% adherence levels, which reduced the mean per-patient annual journey costs to US$340 and US$227, respectively, with corresponding statewide annual investment requirements of US$141 million and US$75 million.

We find significant cost variations across facility types and insurance status. Care costs are higher in private and secondary facilities than in public and primary centres. Since medications account for most expenses, adjusting prescription patterns is key to improving efficiency. Insured patients face modestly higher costs, which may reflect greater adherence to care as well as differences in prescribing patterns. Policies promoting prescription of affordable, quality-assured generics over expensive branded drugs are needed to control DM care costs.

Our results here are comparable to and reinforce the findings from a similar study earlier conducted by our team on hypertension care using patients drawn from the same Banigbe²² used in the present study and the same reference resource price data. The paper reported an average investment gap of US$120 per patient and a statewide annual investment requirement of about US$300 million to implement complete care journeys for all patients with hypertension, translating to US$5000 to US$13 000 per saved life year.³⁸ Collectively, these studies highlight deficiencies in care for NCDs generally and the substantial resources needed to optimise this care in Lagos State.

In our study sample, 5% received insulin therapy. Poor access to insulin has been recognised as a major challenge to management of diabetes in LMICs, and the low 5% insulin use in our study may, at least in part, reflect this.⁴⁹ While country differences exist, estimates from Basu et al suggest that an average of 12.7% of patients with type 2 DM in sub-Saharan Africa will receive insulin therapy in an ideal scenario of universal access to therapy, compared with the present average use of 1.8% with limited access to insulin.⁴⁹ The latest report by the Access to Medicines Foundation showed only 29 of the 108 countries scoped have all the insulins from WHO’s essential medicines list registered and only one of those is a low-income country while no insulins were found to be registered at all in 24 countries.⁵⁰ Increasing access to human as well as analogue insulins should be a global health priority to foster universal access to high quality diabetic care.

Our regression analysis identified frequency of consultations, health insurance coverage, age, and insulin use as significant determinants of blood glucose control—findings that are consistent with broader evidence from LMICs.^{51 52} Regular consultations are often constrained by OOP costs and limited health workforce availability leading to poor continuity of care. Insurance coverage reduces financial barriers and enhances treatment adherence, but coverage levels remain low in Nigeria and many LMICs. Insulin use was associated with poorer control in our cohort, implying that, on average, patients on insulin had higher blood sugar compared with those not taking insulin. This associative finding must not be interpreted to mean that insulin causes a rise in sugar levels. It rather reflects the conventional clinical practice of prescribing insulin for patients with more advanced diseases or poorly controlled sugar. Age-related differences suggest younger patients may face additional challenges with adherence, potentially linked to occupational, social, or cultural barriers. While the evidence on the relationship between gender and glycaemic control has been mixed, our finding of no relationship between gender and sugar control aligns with studies that conclude that gender alone is not a strong predictor, but interacts with socioeconomic, cultural and behavioural factors.⁵¹ Overall, these determinants highlight the need for system-level interventions to address the multilayered barriers to effective glycaemic control in LMIC contexts.

Policy implications

Our findings indicate a critical underfunding for diabetes care and NCDs generally in Lagos. The high prevalence of uncontrolled blood glucose levels and the significant cost burden of medications highlight substantial gaps in the current care and funding landscape. Enhanced financial support from both governmental and non-governmental sources is essential therefore to improve healthcare infrastructure, provide necessary medications, and support comprehensive diabetes management programmes. Yet, we also calculate that meeting the funding requirement for optimum care by direct financial investment alone is not efficient. This necessitates the exploration of more efficient (1) medicine procurement processes and (2) cheaper DM care delivery models.

Medicines, constituting most of the care costs, present a critical point for cost-reduction interventions. First, we recommend bulk purchasing practices to leverage economies of scale and reduce the per-unit cost of medications making them more affordable for patients.⁵³ Shared procurement platforms, possibly coordinated at a regional or national level, are strongly encouraged to enhance purchasing power and streamline the supply chain.^{54 55} Further, the government should engage in strategic negotiations with pharmaceutical companies to lower prices while promoting the use of generic drugs. Ultimately, the government must prioritise enhancement of the local production capacity for diabetes medications and other healthcare resources. In this regard, the recently launched Presidential Initiative for Unlocking the Healthcare Value Chain (PVAC)⁵⁶ which aims to boost local production of pharmaceuticals and medical supplies in Nigeria is a welcome initiative.⁵⁷

Beyond recommending policies that promote the prescription of affordable, quality-assured generics over expensive branded drugs, there is also a need to incentivise physicians toward appropriate prescribing practices. Strategies may include: (1) integrating clinical decision support systems into prescribing platforms to guide evidence-based choices (where electronic prescription platforms exist); (2) incorporating rational prescribing into performance-based financing and value-based care models; (3) linking insurance reimbursements to adherence to national and state treatment guidelines and (4) providing continuous medical education and peer-review feedback on prescribing patterns. Such interventions could help align physician behaviour with cost-effective diabetes management while safeguarding patient outcomes.

We advocate for the implementation of more efficient service delivery approaches, such as group therapy sessions.⁵⁸ Group therapy provides peer support and learning and fosters a supportive community environment that can improve patient adherence to treatment regimens⁵⁸ while reducing the strain on healthcare providers by allowing them to manage multiple patients simultaneously. Additionally, governments at all levels must recognise prevention as a cornerstone of diabetes care policy to reduce the incidence of diabetes. Public health initiatives aimed at promoting healthy lifestyles, such as regular physical activity, balanced diets and weight management, and early detection through targeted screening programmes have been shown to be not only effective but cost-effective and cost-saving.⁵⁹

Our study highlights the need for innovative (digital) health solutions to enhance diabetes care efficiency and effectiveness in Lagos State. From a health system perspective, we recommend investments in digital platforms such as electronic health records that have the capacity to improve medication targeting and evidence-based prescription practices and track patient adherence to follow-ups,⁶⁰ as well as tools linking providers and patients to digital drug supply chain systems to streamline procurement, ensuring consistent medication availability and better affordability.⁵⁵ Equally recommended are mobile health apps that can empower patients with personalised reminders, glucose monitoring support and health education, fostering proactive self-management.⁶¹

The findings underscore the urgent need for policymakers to intensify health insurance expansion interventions in the State. Health insurance schemes should be designed to cover essential diabetes care services and medications, thereby reducing OOP expenses for patients and improving overall health outcomes. Considering the documented evidence of disproportionate economic burden of diabetes care on the poorest populations, policymakers should urgently operationalise the Vulnerable Group Fund to cover the most vulnerable patients, as stipulated in the 2022 National Health Insurance Act of Nigeria.

Lastly, expanding fiscal space for NCD care in Nigeria is crucially recommended to manage the rising economic burden. Beyond increasing health insurance and budget allocation, innovative financing like sin taxes should be explored. The existing 10 Naira per litre tax policy on sugary drinks is short of constitutional ring-fencing for health. An earlier political economy analysis showed strong pro-NCD momentum, which NCD advocates can leverage for legislative amendments to allocate these funds for NCD care.⁶² Raising the tax to the WHO-recommended 20% ad valorem could further support prevention and treatment while discouraging unhealthy consumption.

Strengths and limitations

This study leverages a comprehensive dataset from diverse LSHS facilities, ensuring a robust patient sample. Using actual patient records and price data enhances realism and generalisability, avoiding modelling limitations. The longitudinal approach captures treatment changes over time, refining care and cost analysis. Mixed-effects regression provides nuanced insights into blood glucose control, considering individual and facility-level variations.

This study has some limitations. First, by excluding patients under 30, we aimed to capture type 2 diabetes; however, some adults with long-standing or late onset type 1 diabetes may still be included, which could introduce misclassification bias. On the flip side, we could have missed some young-onset type 2 diabetes patients in the population (if any). Lack of data on the prevalence of these categories of patients in this setting limits our understanding of the extent to which this is a problem for our analyses.¹ The retrospective data collection and reliance on health facility records may introduce biases due to incomplete or inaccurate records. Clinical data were from 2019, while cost data were based on 2023 prices. Although key variables were adjusted for, unmeasured factors such as socioeconomic status, medication adherence and lifestyle may influence outcomes. Some patients may have visited other facilities or used unrecorded treatments. The baseline visit was set as the first visit during the study, though many patients had been receiving care earlier. The time of initial diagnosis or first-ever visit was unknown and not included in the analysis. Findings are specific to Lagos and may not fully apply to regions with different healthcare systems. Moreover, by excluding patients with complications, we limit the generalisability of the findings to the broader DM population. Our extrapolation approach assumes structural similarity between the study cohort and the broader adult diabetic population, but this may not fully hold due to selection and exclusion criteria. Our extrapolation results therefore must be taken for what it is—indicative, rather than inferential, estimates to illustrate potential system-level implications. Lastly, the risk evaluation derives from Bolden-Albala et al.³⁷ This is applicable to our study to the extent that it provides evidence of the risks of cardiac events and strokes in diabetics with controlled and uncontrolled sugar over a 7-year period, using a mixed population that includes patients of black race. However, its applicability is substantially limited by differences in the study population biology, lifestyles, socioeconomic profiles and broader healthcare system designs.

Conclusions

This study highlights a significant gap between the current and optimal diabetes care in Lagos State, underscoring the critical need for improved diabetes management strategies, including increased frequency of consultations and enhanced patient education and adherence support to achieve optimal blood glucose control. There are substantial potential health and economic benefits from such optimisation of diabetes care in the State, but realising these benefits will require substantial additional investment alongside efficiency gains in medicines procurement and care delivery. There is a clear need to explore innovative (digital) provider and patient journey schemes such as group therapy to incentivise patient adherence to (primary and secondary) prevention behaviours and reduce care costs. Moreover, substantial reductions in DM medicines prices are required and this could be addressed by (combinations of) pooled and bulk drug procurement, increased emphasis on generic medications prescription, implementation of differential DM medicine pricing for LMICs, and investments in local DM medicines production capacity while expanding health insurance coverage.

Data availability statement

Data are available on reasonable request.