Health Insurance and Health Utilization in Uasin Gishu County, Kenya

Health Insurance and Health Utilization in Uasin Gishu County, Kenya

Tanui Kiprotich Leonard, Dr. Issacs Kemboi, Dr. Simeon Nganai

Department of Economics, Moi University, Kenya

DOI: https://dx.doi.org/10.47772/IJRISS.2025.909000460

Received: 15 September 2025; Accepted: 22 September 2025; Published: 15 October 2025

ABSTRACT

This paper empirically examines the influence of health insurance on health care utilization in Uasin Gishu County, Kenya. The study adopted an explanatory research design and employed a quantitative cross-sectional household survey. Data were collected using structured questionnaires from residents sampled proportionally across the county’s six constituencies. Hypotheses were tested using binary logistic regression, and average marginal effects were estimated to translate coefficients into absolute probability differences. With utilization coded as any use in the recall period, insurance showed a statistically significant negative association with care-seeking, indicating that nominal coverage did not translate into higher realized use in this setting. Taken together, the findings point to implementation frictions, benefit awareness gaps, residual point-of-service charges, and supply-side constraints as plausible mechanisms that blunt insurance’s expected enabling effect. County health managers should therefore clarify entitlements and eliminate informal fees; assure availability of essential medicines and basic diagnostics at first contact; realign empanelment and contracting to where clients actually seek care; and expand time-compatible access through after-hours/weekend clinics, express lanes, workplace/mobile outreaches, and appointment systems. Embedding assisted enrollment/renewal and monitoring insured visits by sector can further convert coverage into first contact.

Keywords: Health care utilization; Health Insurance; Logistic regression; Uasin Gishu County, Kenya.

INTRODUCTION

Universal health coverage (UHC) remains a central health and development priority worldwide, anchored in Sustainable Development Goal (SDG) 3.8. Yet the most recent joint monitoring indicates progress has stalled. The 2023 Global Monitoring Report on UHC shows service coverage has largely stagnated since 2015, while financial protection has worsened: more than one billion people about 14% of the world’s population face catastrophic out-of-pocket (OOP) health spending exceeding 10% of household budgets. Globally, an estimated 4.5 billion people still lack full access to essential health services, underscoring the gap between aspiration and lived reality. These trends highlight the dual UHC mandate: expand service coverage and protect households from financial hardship. Health insurance public, national, or community-based is consistently framed as a key lever for achieving both aims.

A substantial body of evidence links health insurance to increased health-care utilization in low- and middle-income countries (LMICs)2w, although effects on financial risk protection and quality can be mixed. A frequently cited systematic review across LMICs found that public health insurance generally increases utilization of both inpatient and outpatient care and can improve financial protection, though results vary by design and context (e.g., benefit packages, provider payment methods, and supply-side readiness) (Erlangga et al., 2019). More recently, a 2023 PLOS ONE meta-analysis focusing on community-based health insurance (CBHI) in LMICs also reported significant gains in service utilization, affirming insurance’s demand-side role in pulling individuals into care. At the same time, reviews and policy analyses caution that without parallel investments in service quality and system capacity, insurance-induced demand may not translate into better outcomes.

Across Sub-Saharan Africa, the UHC picture mirrors global headwinds. WHO’s Africa Region tracking notes stagnation or even slight declines in the UHC Service Coverage Index between 2019 and 2021, reflecting persistent bottlenecks in primary health care (PHC), constrained public financing, and lingering COVID-19 shocks. Many countries still rely heavily on OOP payments, and insurance coverage remains patchy, especially among informal-sector households. As a result, improving prepayment and risk pooling (through social, national, or community-based schemes) and strengthening PHC networks are repeatedly identified as core strategies for African health systems seeking to raise utilization without exacerbating financial hardship.

Kenya has, over the last decade, explicitly pursued UHC through health-financing and service-delivery reforms. The UHC Policy 2020–2030 prioritizes PHC as the first point of contact and emphasizes shifting from OOP to pooled financing mechanisms. In 2023, Kenya enacted the Social Health Insurance Act (SHIA), establishing the Social Health Authority (SHA) and the Social Health Insurance Fund (SHIF) to replace the National Hospital Insurance Fund (NHIF) and to modernize contributions, benefits, and governance. Government communications and regulatory instruments issued in late 2023–2024 outline the transition architecture and the roll-out of new benefits and tariffs, with implementation staged from October 2024. In parallel, the Ministry of Health (MoH) is scaling primary care networks (PCNs) to reorganize service delivery around connected PHC units, with updates through 2024 reporting substantial progress across counties. Together, these policy shifts aim to widen coverage, stabilize financing, and improve access and continuity of care—conditions expected to raise appropriate utilization.

Despite these advances, baseline coverage data show room to grow. The Kenya Demographic and Health Survey (KDHS) 2022 reports that about one in four Kenyans is covered by some form of health insurance (26% of females and 27% of males), with NHIF the most common type (~24% among both sexes). Coverage is notably higher in urban settings (~39% of females, 41% of males) than rural areas (~20% and 19%), and it rises sharply with wealth. These gradients indicate persistent geographic and socioeconomic inequities that can blunt the effect of insurance reforms if not addressed with targeted outreach and subsidization. preview.dhsprogram.com

At the same time, national surveys on health expenditure and utilization emphasize the stakes. The Kenya Household Health Expenditure and Utilization Survey (KHHEUS) 2018—the latest fully published round documents patterns of care-seeking, OOP spending, and insurance penetration at national and county levels, providing a baseline for monitoring UHC reforms. Analyses drawing on KHHEUS data have shown inequalities in utilization and financial risk, particularly affecting poorer households; such disparities can dampen improvements in population-level utilization even when aggregate coverage rises.

Within this national landscape, Uasin Gishu County occupies a distinctive position. Economically dynamic and strategically located in Kenya’s Rift Valley, the county hosts the Moi Teaching and Referral Hospital (MTRH)—the national referral hub for western Kenya and a dense network of public facilities. A 2024 study conducted in Uasin Gishu describes 125 public health facilities, including one national referral hospital, two district hospitals, 11 sub-district hospitals, 23 health centres, and 88 dispensaries, underscoring substantial service supply for both county residents and referrals from neighboring regions. This supply-side capacity especially with a tertiary referral anchor can facilitate utilization if financing and navigation barriers are minimized.

Local evidence also sheds light on how insurance affects real-world access and use in Uasin Gishu. A 2024 BMC Health Services Research paper examined NHIF-capitated members in the county and found that four factors strongly predicted access to PHC services: (i) patient knowledge of the benefit package, (ii) premium payment processes, (iii) selection of the capitated facility, and (iv) the effectiveness of NHIF communication. Notably—and problematically—80% of respondents reported paying out-of-pocket at some point for items like registration, consultation, medicines, or laboratory tests despite being capitated, which can deter timely care-seeking and reduce the utilization gains expected from insurance. These insights point to both demand-side (information, affordability, and expectations) and supply-side (availability of covered services and medicines) barriers that shape utilization under social health insurance.

County planning documents echo these concerns. The Uasin Gishu County Integrated Development Plan (CIDP) 2023–2027 lists health as a priority sector and, within its social protection agenda, targets incremental NHIF enrollment of vulnerable residents (e.g., 1,000 beneficiaries per year) to expand coverage among those least able to pay. Such county-level actions are intended to complement national reforms—transitioning from NHIF to SHIF and consolidating PCN-based service delivery—by pushing coverage deeper into underserved groups and, in turn, raising appropriate utilization of outpatient and preventive services.

From a health-systems perspective, the mechanisms by which insurance influences utilization in Kenya and Uasin Gishu are consistent with international literature but are also highly context-dependent. Insurance lowers the price at the point of service, reducing financial barriers to seeking care earlier and more frequently, particularly for outpatient visits and chronic-disease management. Studies across Africa and Asia including recent cross-country analyses that encompass Kenya link insurance coverage to higher odds of facility delivery, antenatal care, childhood immunization, and treatment-seeking for common illnesses, although effects vary with benefit depth and provider payment methods. Where capitation is used, as in Kenya’s primary care, there can be complementary incentives for providers to emphasize prevention and continuity—but only if drug availability, staffing, and referral pathways are reliably financed and managed. Thus, as Kenya’s SHIF and PCN reforms mature, utilization gains will likely hinge on aligning purchasing arrangements with service-delivery capacity while minimizing co-payments and eliminating informal charges that erode trust.

For Uasin Gishu County, several features heighten the importance of studying health insurance and utilization now. First, the policy transition from NHIF to SHIF will change contribution rules, benefit designs, and administrative processes; how households and providers adapt may directly affect primary care attendance, referrals to MTRH, and the balance between public and private providers in Eldoret’s mixed market. Second, the county’s PCN roll-out and status as a regional referral hub create an opportunity to evaluate whether insurance-enabled PHC can decongest tertiary services by improving access at lower-level facilities. Third, persistent inequalities in coverage (urban–rural, wealth-related) documented in KDHS 2022 are plausibly mirrored within the county, suggesting that insurance-driven utilization may be uneven unless accompanied by targeted enrollment and supply-side strengthening in underserved wards.

Literature review and Hypotheses development

Sikka et al. (2021) analyze baseline data from 1,447 adults with elevated blood pressure in the LARK Hypertension study (Kosirai and Turbo divisions in western Kenya) using latent class regression to classify utilization and costs; they find women report more outpatient visits and prescriptions than men (39% vs 28% and 42% vs 30%) and have higher odds of belonging to a high-cost utilizer class alongside worse self-reported health, highlighting persistent sex gaps in both service use and financial burden in rural western Kenya (including Uasin Gishu’s Turbo Division).  Dowden et al. (2019) evaluate “male clinics” in rural western Kenya—monthly sessions with male providers and male-friendly spaces within existing facilities—and show significant increases in male attendance and preventive “check-up” visits (≈20% of visits in intervention vs ≈1% in controls), suggesting that demand-side barriers rooted in gender norms and clinic design can be reduced through service reframing, with implications for broader utilization patterns under insurance schemes.

Wambiya et al. (2021) conduct a cross-sectional study in a Nairobi informal settlement and, using multinomial models, show that having health insurance is strongly associated with using private facilities (aRRR≈2.95) while cost satisfaction also pushes clients to public facilities; overall, 47% sought care in private facilities vs 33% in public, underscoring how insurance and affordability shape pathways into care among the urban poor.  Were, Mwangi, and Muiruri (2024) focus on Uasin Gishu County and find that among NHIF-capitated members, access to primary care hinges on members’ knowledge of the benefit package, communication from NHIF, premium payment processes, and facility selection rules; despite prepayment, 80% reported some out-of-pocket spending, indicating information and payment-process frictions that blunt insurance-driven utilization gains.

Shearer et al. (2021) analyze surgical admissions at a tertiary hospital in Eldoret and report gender differences in NHIF enrollment and health-seeking behaviors around surgical care, implying that prepayment prerequisites for high-cost admissions can differentially delay or facilitate access by sex, with downstream consequences for utilization intensity and outcomes.  Seidu (2020) pools DHS data from 24 sub-Saharan African countries (n≈308,000 women) and shows that lack of insurance, rural residence, low wealth, and manual occupations significantly increase the odds of reporting barriers to care; conversely, insurance coverage is associated with lower barriers, linking coverage to realized access and potential utilization gains.

Ilinca et al. (2019) decompose Kenya’s KHHEUS-2018 and find pro-rich inequities in outpatient, inpatient, and preventive care use; education and living standards are dominant drivers of inequality while age and gender also contribute, indicating that insurance expansion must be coupled with equity-sensitive purchasing to shift utilization toward need. Abajobir et al. (2024) run a cluster randomized trial in Kakamega County testing digital nudges to catalyze NHIF uptake for RMNCH services; the intervention increased insurance coverage and improved selected service-use indicators, demonstrating that targeted enrollment supports can translate insurance eligibility into actual utilization along the maternal-child continuum.

Kazungu et al. (2024) analyze patient choice and provider competition under NHIF and report that plan design and contracting shape where insured Kenyans seek care, with implications for quality and responsiveness; strategic purchasing (e.g., accreditation, capitation rates) can steer insured utilization and stimulate competitive effects. Mbau et al. (2020) examine NHIF’s 2015 reforms (premium, benefits, payment methods) and conclude that weaknesses in design and implementation limited strategic purchasing, with negative implications for equity, efficiency, and quality—factors that condition whether insurance translates into expanded and appropriate utilization.

Obadha et al. (2019) study provider experiences with NHIF capitation vs fee-for-service in Kenya and find that capitation’s administrative simplicity coexists with concerns about rates and perceived under-provision risks; provider responses to payment method can influence availability, referral behavior, and thus insured patients’ realized use. Erlangga et al. (2019) synthesize 68 studies from LMICs and conclude that public health insurance generally increases service utilization and financial protection, with mixed but often positive effects on health status—reinforcing the expectation that expanding NHIF-like coverage should raise contacts with care when other barriers are managed.

Eze et al. (2023) meta-analyze community-based health insurance (CBHI) across LMICs and find significant increases in outpatient visits and care-seeking plus reductions in catastrophic spending among members, indicating that even voluntary, small-scheme coverage can move the utilization needle for poorer populations.  Osei et al. (2025) analyze Ghana’s 2022 DHS and observe that insurance is associated with higher odds of postnatal care, but paradoxically lower odds of facility delivery and skilled birth attendance—pointing to implementation gaps (e.g., residual out-of-pocket charges, reimbursement delays) that can dampen utilization effects despite nominal coverage.

Nyeri County’s UHC pilot study (2023) reports improvements in hospital performance and efficiency metrics after user-fee removal and pooled purchasing, consistent with increased service volumes and more appropriate resource use—evidence that financing reform can quickly change utilization patterns at facility level.  A national implementation assessment of Kenya’s UHC pilots (2024) documents organizational and financing challenges (governance, contracts, benefit clarity) that mediated the pilots’ effects on access and use, emphasizing that insurance expansion must be paired with strong purchasing and provider-payment reforms to sustain utilization gains.

The 2022 Kenya Demographic and Health Survey (KDHS) reports that only about one in four Kenyans has any health insurance (26% females; 26.5% males), with NHIF the dominant scheme (24% each) and stark gradients by residence (≈39–41% urban vs ≈19–20% rural) and wealth—structural coverage gaps that constrain utilization among poorer and rural households. Using KHHEUS-2018, Salari et al. (2019) estimate that out-of-pocket payments push roughly 1.0–1.1 million Kenyans into poverty, with catastrophic payments concentrated among the poorest and driven largely by outpatient care; without effective prepayment and purchasing, cost pressures can suppress needed utilization or drive ruinous spending.  Mtei et al. (2023) assess Tanzania’s improved Community Health Fund (iCHF) and show that insured households are more likely to seek formal outpatient care and have fewer foregone-care episodes than the uninsured, indicating that sub-national prepayment schemes can raise utilization when benefits are credible and providers contracted.  Damtie et al. (2024) synthesize Ethiopia’s CBHI evaluations and find consistent increases in modern care use and reductions in out-of-pocket spending among enrollees, with effects amplified where benefit packages are transparent and renewal/enrollment is simplified mechanisms relevant for Kenya’s transition to new social health insurance. Based on the literature reviewed above, this research hypothesize as follows:

Ho1: Health insurance significantly influences Health-Care Utilization in Uasin Gishu County

Target population and data

The target population for this study comprised all residents of Uasin Gishu County, Kenya across the six constituencies—Soy, Kapseret, Kesses, Turbo, Moiben, and Ainabkoi—totaling 1,163,186 persons as per the 2019 Kenya Population and Housing Census. In line with the study objectives, the population is stratified by constituency to reflect both urban and rural settlement patterns. Primary data was collected from household respondents (urban and rural) using a structured, self-administered questionnaire, while the KNBS (2019) census figures provide the sampling frame for proportional allocation across strata. Unlike a census, the study employs sampling to obtain a representative cross-section of households from each constituency.

Table 1: Target population

Source: KNBS (2019); computations by Researcher (2025).

Sample size

The study utilized Cochran’s formula to determine the sample size, the formula is  widely accepted in Survey research when the population size is large and the proportion of interest like the health utilizing proportion is unknown or assumed to be 50%. (Cochran in 1977). The formula was given by;

Where:  = required sample size, z= z-value (1.96 for 95% confidence), p = estimated proportion of an attribute present in the population (use 0.5 for maximum variability) and e = desired margin of error (e.g., 5% = 0.05).

Data was analyzed using Stata version 14 using the Binary Logistic Regression, in this model the probability that an individual accessed healthcare services in the past 12 months. is determined, 1 = Accessed healthcare while 0 = Did not access healthcare. A binary logistic regression was conducted to identify the influence of health insurance on the likelihood of utilizing healthcare.

Model specification

Count data faces the problems of overdispersion and excess zeros, to handle this problem, the study used of the hurdle model.

 …………………………………………………………………(1)

The positive counts form the second-density function

……………………………..…(2)

The model collapses to the standard count model only if f1 (0) = f2 (0)

The first density function f1 [.] is normally estimated using logistic regression while the second-density function f2 [.] is estimated using count data models like Poisson or negative binomial model. The mean of the hurdle model is determined by the probability of crossing the threshold and by the moments of the zero-truncated density as follows:

 ……………………………………………………………(3)

Where μ_2 (X) is the untruncated mean in density f_2 (y/X) . The variance of hurdle model is given by;

………….….(4)

Generally, π_i is modeled with a logistic regression and μ_i is modeled as a log-linear regression. The ZI model can be written as,

Log µ_1= (x_1^Tα,Logit (π_1 )=Z_1^T β

Where: α and β are regression coefficients forx_1^T the covariates and Z_1^T β

We assume that the decision to utilize healthcare services is determined by some identified factors (independent variables) as shown below.

U= (β₀ + β₁X₁ + β₂X_{2 ……+} β_nX_n)

VST= β₀+ β₁HI₁ ………………………………………………..(5)

Where: VST is the number of Visits to the hospital. HI is the health insurance, β_1, are regression Coefficients.

Data analysis and interpretation

Goodness of fit test

The Pearson Goodness-of-Fit test is a diagnostic tool commonly used to evaluate how well a logistic regression model fits the observed data. Specifically, it compares the observed frequencies of outcomes to the expected frequencies predicted by the model. In this case, the test yielded a Chi-square value of 52.87 with 51 degrees of freedom, and a corresponding p-value of 0.4016.

Since the p-value is greater than the conventional alpha level of 0.05, we fail to reject the null hypothesis that the model fits the data well. This implies that there is no significant difference between the observed and expected values, suggesting that the model has an adequate fit (Hosmer, Lemeshow, & Sturdivant, 2013).

In practical terms, this result indicates that the logistic regression model does not suffer from major specification issues, such as omitted variables or incorrect functional form, at least with respect to the distribution of the predicted probabilities. Therefore, the model’s predictive accuracy and parameter estimates can be considered reliable for interpreting the relationship between the independent variables and the binary outcome.

Table 2: Goodness-of-Fit Test (Pearson Chi-Square)

Source: Field data (2025)

Model Information Criteria

The Model Information Criteria are essential for evaluating the fit and parsimony of a logistic regression model. These metrics are particularly useful for model comparison and selection. First, the Log Likelihood (null) value of -158.086 represents the fit of a baseline model that includes only the intercept and no predictor variables. In contrast, the Log Likelihood (model) value of -145.637 corresponds to the fitted model that includes the independent variables. The less negative log-likelihood in the full model indicates improved model fit compared to the null model, meaning that the inclusion of predictors significantly enhances the model’s ability to explain the outcome (Menard, 2002).

The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) provide further insight. The AIC value of 301.27 and BIC value of 318.84 are both penalized-likelihood criteria that balance model fit with model complexity. Lower values of AIC and BIC indicate better models, especially when comparing multiple competing models (Burnham & Anderson, 2004).

In this case, the reported AIC and BIC suggest a reasonable trade-off between goodness-of-fit and overfitting. However, these values are most informative when used to compare this model against alternative specifications; on their own, they do not convey an absolute measure of quality. Nonetheless, the improvement in log-likelihood from the null to the full model, along with the relatively moderate AIC and BIC values, supports the notion that the model is statistically robust and appropriately specified.

Table 3: Model Information Criteria

Source: Field data (2025)

Regression results

The purpose of this analysis was to assess whether health insurance predicts health-care utilization defined here as a binary outcome with 1 = used care in the recall period and 0 = did not use in Uasin Gishu County using a binary logistic regression framework. In the Behavioral Model of Health Services Use, insurance is conceptualized as an enabling resource that should reduce financial barriers and increase realized care-seeking when need is present (Andersen, 1995). Following best practice in applied logistic regression, we interpret the model coefficient on both the relative scale—by exponentiating the log-odds coefficient to obtain an odds ratio (OR = e^β)—and the absolute scale—by reporting average marginal effects (AMEs; “margins”) that translate results into percentage-point changes in predicted probability (Hosmer, Lemeshow, & Sturdivant, 2013; Williams, 2012). Statistical decisions are based on α = .05.

In the fitted logit model (Table 4), the coefficient for Health insurance is β = −0.739 (SE = 0.327, p = .024; 95% CI: −1.38, −0.098). With the outcome coded as utilization = 1, the negative sign indicates that insured individuals have lower odds of using care than the uninsured, holding other model terms constant. Converting to the relative scale yields OR = e^−0.739 ≈ 0.48, implying about a 52% reduction in the odds of any use among the insured; the 95% CI for the OR is 0.25 to 0.91, derived from exponentiating the coefficient’s confidence limits. To complement this with an absolute measure, the margins output (Table 5) shows AME = − 0.149 (SE = 0.063, p = .019), meaning that moving from uninsured to insured is associated with a 15–percentage-point lower probability of using care on average, across the observed covariate distribution (Williams, 2012). These quantities are internally consistent with the model intercept: the constant β₀ = 1.314 (p = .002) corresponds to a baseline predicted probability of ~0.79 for uninsured respondents, while adding the insurance coefficient (1.314 − 0.739 = 0.575) yields a predicted probability of ~0.64 for insured respondents — an absolute difference of about −0.15.

Interpreted literally with utilization = 1, this pattern is counterintuitive relative to the broader literature from low- and middle-income countries, which generally reports that public or community-based insurance increases the likelihood of service use and reduces financial hardship (Erlangga, Suhrcke, Ali, & Bloor, 2019; Eze, Ilechukwu, & Lawani, 2023). In Kenya specifically, descriptive evidence shows persistent shortfalls in coverage and strong socioeconomic gradients in both coverage and use, but the direction of association between insurance and use is typically nonnegative once barriers are controlled (Kenya National Bureau of Statistics [KNBS], Ministry of Health [MoH], & ICF, 2023). Several empirically plausible mechanisms can reconcile the present negative association with utilization. First, the measure of “use” may emphasize a subset of providers (e.g., public facilities only), while insured clients substitute toward accredited private or faith-based providers; in such cases, observed “public use” falls despite stable or higher overall use (Wambiya, Otieno, Mutua, Donfouet, & Mohamed, 2021). Second, there may be implementation frictions — for example, residual out-of-pocket charges, medicine stock-outs, or unclear benefit entitlements — that deter insured patients from seeking care even when nominally covered; such gaps have been documented nationally and within Uasin Gishu among NHIF-capitated members (KNBS et al., 2023; Were, Mwangi, & Muiruri, 2024; Mbau, Kabia, Chuma, & Barasa, 2020). Third, the estimate can reflect selection and confounding: healthier or time-constrained formal workers may be more likely to hold insurance and less likely to need or make visits within a short recall window, biasing the unadjusted association downward if key covariates are omitted (Hosmer et al., 2013; Wooldridge, 2010). Fourth, measurement error — for instance, classifying lapsed members as insured or recall bias in utilization — can attenuate or invert effects.

These results therefore highlight both substantive and methodological implications. Substantively, if insurance is truly associated with a ~15-percentage-point lower probability of any use, the policy signal is to examine the user experience of “cashless” care: benefit awareness, facility choice rules, point-of-service fees, and drug availability. Qualitative accounts and county-level surveys suggest that even insured clients may face registration fees, co-payments, or stock-related purchases, which can deter early contact and erode trust — especially for routine primary care (Were et al., 2024; KNBS et al., 2023). Methodologically, transparent reporting should present both the odds ratio (≈0.48) and the AME (≈−0.15) alongside predicted probabilities for insured versus uninsured groups, since stakeholders often find absolute differences more actionable (Williams, 2012). If the current specification is bivariate or minimally adjusted, extending to a multivariable model with predisposing and enabling factors — age, sex, education, wealth, and urban/rural residence — will help address confounding known to be important in Kenya (KNBS et al., 2023). Given potential endogeneity of insurance status, propensity-score weighting/matching or doubly robust estimators can further stabilize causal interpretation, and sensitivity checks to alternative definitions of “use” (any provider vs. public-only) can test for substitution effects (Austin, 2011; Wooldridge, 2010).

Table 4: Logit regression results

Source: Field data (2025)

Table 5: logistic regression margins

Source: Field data (2025)

CONCLUSION AND RECOMMENDATION

This study assessed whether health insurance predicts health-care utilization defined as any use in the recall period in Uasin Gishu County using a binary logistic regression framework grounded in Andersen’s Behavioral Model, which treats insurance as an enabling resource that should lower financial barriers and convert perceived need into realized care (Andersen, 1995). Interpreted on both relative and absolute scales, the analysis indicates a statistically significant negative association between insurance status and utilization when coded this way, a result that is counterintuitive relative to much low- and middle-income country evidence linking coverage to higher service use and better financial protection (Erlangga, Suhrcke, Ali, & Bloor, 2019; Eze, Ilechukwu, & Lawani, 2023). Plausible mechanisms in the Kenyan context include implementation frictions—residual point-of-service fees, medicine stock-outs, and unclear benefit entitlements—alongside provider substitution (insured clients shifting toward accredited private or faith-based facilities not captured by the measure of “use”), selection and confounding from omitted socio-demographic factors, and potential measurement error in insurance status or utilization recall (Mbau, Kabia, Chuma, & Barasa, 2020; KNBS, MoH, & ICF, 2023; Wambiya, Otieno, Mutua, Donfouet, & Mohamed, 2021; Hosmer, Lemeshow, & Sturdivant, 2013). Accordingly, recommendations are to clarify entitlements and eliminate informal charges; secure essential medicines and basic diagnostics at first contact; realign empanelment and contracting with where clients actually seek care; expand time-compatible access via after-hours/weekend clinics, express lanes, workplace/mobile outreaches, and appointment systems; embed assisted enrollment and renewal with transparent claim confirmations; and monitor insured visits by sector to detect substitution. For analytic robustness and clearer policy translation, report odds ratios and average marginal effects with predicted probabilities, adjust for key predisposing/enabling covariates, and mitigate selection using propensity-score or doubly robust approaches, with sensitivity analyses that define “use” as care from any provider (Williams, 2012; Austin, 2011; Wooldridge, 2010).

REFERENCES

1. Abajobir, A., de Groot, R., Wainaina, C., Pradhan, M., Janssens, W., & Sidze, E. M. (2024). The impact of digital interventions on health insurance coverage for reproductive, maternal, newborn and child health services utilization in Kakamega, Kenya: A cluster randomized controlled trial. Health Policy and Planning, 39(10), 1007–1021. (context on Kenya NHIF interventions).
2. Andersen, R. M. (1995). Revisiting the behavioral model and access to medical care: Does it matter? Journal of Health and Social Behavior, 36(1), 1–10.
3. Austin, P. C. (2011). An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behavioral Research, 46(3), 399–424. https://doi.org/10.1080/00273171.2011.568786
4. County Government of Uasin Gishu. (2022). County Integrated Development Plan (CIDP) 2023–2027. Eldoret: CGUG. (health and social protection targets, including NHIF enrollment).
5. Damtie, Y., et al. (2024). Community-based health insurance and healthcare utilization in Ethiopia: A systematic review and meta-analysis. [Journal/Publisher]. (Advance online publication).
6. Dowden, J., Mushamiri, I., McFeely, E., Apat, D., Sacks, J., & Ben Amor, Y. (2019). The impact of “male clinics” on health-seeking behaviors of adult men in rural Kenya. PLOS ONE, 14(11), e0224749.
7. Erlangga, D., Suhrcke, M., Ali, S., & Bloor, K. (2019). The impact of public health insurance on health care utilisation, financial protection and health status in low- and middle-income countries: A systematic review. PLOS ONE, 14(8), e0219731. (evidence of utilization gains under insurance).
8. Eze, P., Ilechukwu, S., & Lawani, L. O. (2023). Impact of community-based health insurance in low- and middle-income countries: A systematic review and meta-analysis. PLOS ONE, 18(6), e0287600. (evidence of increased utilization under CBHI).
9. Health Policy Open. (2023). The effect of a pilot universal health coverage program on hospital performance in Nyeri County, Kenya. Health Policy Open.
10. Health Systems & Reform. (2024). Examining the implementation of Kenya’s UHC pilot. Health Systems & Reform. https://doi.org/10.1080/23288604.2024.2418808.
11. Hosmer, D. W., Lemeshow, S., & Sturdivant, R. X. (2013). Applied logistic regression (3rd ed.). Wiley.
12. Ilinca, S., Di Giorgio, L., Salari, P., Chuma, J., & Kwesiga, B. (2019). Socio-economic inequality and inequity in healthcare use in Kenya. International Journal for Equity in Health, 18, 196.
13. Kazungu, J., Obadha, M., Kinyenje, A., & Barasa, E. (2024). Patient choice and provider competition under Kenya’s NHIF. BMC Health Services Research, 24, 853.
14. KEMRI-Wellcome Trust Research Programme. (2024). Examining the implementation experience of Primary Care Networks in Kenya. Nairobi: KEMRI-Wellcome. (service-delivery reforms supporting utilization).
15. Kenya Ministry of Health. (2023). Social Health Insurance (General) Regulations. Nairobi: MoH. (framework implementing the Social Health Insurance Act and establishing SHA/SHIF).
16. Kenya Ministry of Health. (2024). Kenya makes strides in Primary Health Care Network implementation (MoH update). Nairobi: MoH. (PCN progress to support service access).
17. Kenya National Bureau of Statistics (KNBS), Ministry of Health (MoH), & ICF. (2023). Kenya Demographic and Health Survey (KDHS) 2022: Key Indicators Report. Nairobi and Rockville, MD. (national health insurance coverage statistics, urban–rural and wealth gradients).
18. Kenya National Bureau of Statistics (KNBS), Ministry of Health (MoH), & ICF. (2023). Kenya Demographic and Health Survey 2022: Key Indicators Report. KNBS.
19. Mbau, R., Kabia, E., Chuma, J., & Barasa, E. (2020). Examining purchasing reforms towards UHC by the NHIF in Kenya. International Journal for Equity in Health, 19, 58.
20. Ministry of Health & Kenya National Bureau of Statistics. (2018). Kenya Household Health Expenditure and Utilization Survey (KHHEUS) 2018. Nairobi: MoH & KNBS. (national baseline on utilization, OOP, insurance).
21. Obadha, M., Chuma, J., Kazungu, J., & Barasa, E. (2019). Health care purchasing in Kenya: Provider experiences with capitation and fee-for-service. The International Journal of Health Planning and Management, 34(1), e917–e933.
22. Osei, K. M., Prasiska, D. I., Chapagain, D. D., et al. (2025). Health insurance enrollment and maternal health service utilization using Ghana DHS 2022. PLOS ONE, 20(6), e0325240.
23. Salari, P., Di Giorgio, L., Ilinca, S., & Chuma, J. (2019). The catastrophic and impoverishing effects of out-of-pocket healthcare payments in Kenya, 2018. BMJ Global Health, 4(6), e001809.
24. Seidu, A.-A. (2020). Mixed-effects analysis of barriers to healthcare among women in sub-Saharan Africa. PLOS ONE, 15(11), e0241409.
25. Shearer, B., et al. (2021). Gender differences in national insurance enrollment and surgical care in Eldoret, Kenya. Journal of Surgical Research, 267, 72–80.
26. Sikka, N., DeLong, A., Kamano, J., et al. (2021). Sex differences in healthcare utilization and costs among adults with elevated blood pressure: The LARK study from western Kenya. BMC Public Health. (Research article).
27. Tanzania iCHF study. (2023). Factors associated with iCHF enrollment and utilization in Tanzania: Evidence from a mixed-methods assessment. BMC Public Health.
28. Wambiya, E. O., Otieno, P., et al. (2021). Patterns and predictors of private/public healthcare utilization in a Nairobi informal settlement. BMC Public Health, 21, 1117.
29. Were, B. N., Mwangi, E. M., & Muiruri, L. W. (2024). Barriers of access to primary healthcare services by NHIF-capitated members in Uasin Gishu County, Kenya. BMC Health Services Research, 24, 1025. (local evidence on access and OOP among insured).
30. WHO Regional Office for Africa. (2023). Universal Health Coverage in Africa: Highlights (Dec 2023). Brazzaville: WHO AFRO. (regional UHC status and constraints).
31. Williams, R. (2012). Using the margins command to estimate and interpret adjusted predictions and marginal effects. The Stata Journal, 12(2), 308–331.
32. Wooldridge, J. M. (2010). Econometric analysis of cross section and panel data (2nd ed.). MIT Press
33. World Bank. (2023, September 18). Billions left behind on the path to universal health coverage (press release). Washington, DC. (one billion facing catastrophic spending; context for UHC).
34. World Health Organization (WHO) & World Bank. (2023). Tracking Universal Health Coverage: 2023 Global Monitoring Report. Geneva and Washington, DC. (global service coverage stagnation; financial hardship trends).