Digital health capacity and pediatric care quality in LMICs: A large-scale analysis of 5311 health facilities

Abstract

Objective

Digital health technologies are crucial for achieving universal health coverage (UHC), especially in low- and middle-income countries (LMICs) with limited digital infrastructure. This study aimed to assess digital health capacity across multiple LMICs and evaluate its association with evidence-based practice (EBP) and patient-centered care (PCC).

Methods

We analyzed Service Provision Assessment data collected over the past decade, spanning 5311 facilities and 20,880 pediatric visits across eight LMICs. Digital health capacity was measured using the WHO Classification of Digital Health Interventions (CDHI) across five domains: digital infrastructure, client engagement, healthcare providers, data services, and health system managers. EBP was assessed via ten binary items, while PCC was derived from eleven exit interview items using exploratory factor analysis. Multilevel regression models explored relationships between digital health capacity and both EBP and PCC.

Results

Overall digital health capacity was low (mean = 0.35), with notably low scores for digital infrastructure (0.02), healthcare providers (0.21), and health system managers (0.06). Digital health capacity was significantly associated with improved EBP (Coef. = 0.146, p < 0.001), particularly through digital infrastructure (Coef. = 0.183, p = 0.029), client engagement (Coef. = 0.205, p < 0.001), and provider capacity (Coef. = 0.142, p < 0.001). No significant effect emerged for PCC (Coef.=−0.013, p = 0.531).

Conclusions

The level of digital health technology in LMICs is generally insufficient, particularly in terms of digital infrastructure, healthcare provider training and health system managers. Although the implementation of digital health technologies has the potential to improve the EBP, its effect on enhancing PCC is relatively limited.

Introduction

In 2022, an estimated 4.9 million children under five years of age died globally, with those in high-mortality countries facing an 80-fold higher risk compared to their counterparts in low-mortality settings. ¹ Such disparities are closely linked to deficiencies in healthcare quality, limited access to healthcare services, and broader social determinants of health (SDOH)—including economic constraints, educational barriers, and infrastructural limitations—underscoring the need for a comprehensive and equity-focused approach to health system strengthening.^2–5 Enhancing digital health capabilities has emerged as a pivotal strategy in addressing this gap, offering a pathway to achieve the sustainable development goals (SDGs), including universal health coverage, as well as the World Health Organization's “triple billion” targets.^6,7

Digital health represents a transformative approach to healthcare delivery, leveraging information communication technology (ICT), data analytics, and computer science to broaden access, streamline services, and augment health information management. ⁸ Its applications span telemedicine, ⁹ wearable devices, electronic health records (EHRs), and advanced methodologies such as artificial intelligence (AI) and machine learning (ML). ¹⁰ These innovations can bolster key functions in healthcare, including remote service provision, clinical decision support, and patient self-management. ¹¹ The COVID-19 pandemic has further accelerated digital health adoption, particularly in low- and middle-income countries (LMICs), where constraints such as insufficient infrastructure or workforce shortages have prompted creative technological solutions. ¹²

Despite growing interest, the influence of digital health technologies on overall healthcare quality remains inconsistently reported. While a number of studies have demonstrated beneficial outcomes,^12,13 others highlight marginal gains or even adverse effects related to patient safety and service quality.^14–16 Notably, the majority of this literature originates from high-income countries, where digital health has improved chronic disease management and care coordination.^17,18 However, evidence on pediatric healthcare quality in LMICs is comparatively sparse. Although some efforts have been made to reduce inequities,^19,20 persistent challenges include economic barriers, deficient infrastructure, and policy obstacles, ²¹ compounded by limited training for healthcare professionals, weak stakeholder collaboration, and insufficient regulatory frameworks.^22,23 Such hurdles are evident in regions like sub-Saharan Africa and parts of the Middle East, exemplifying the urgent need for robust, context-specific digital health solutions.

This study aims to provide a comprehensive evaluation of digital health capacity in LMICs, particularly their impact on pediatric healthcare quality and user experience. We hypothesize that the implementation of digital health technologies will significantly enhance the efficiency and quality of pediatric healthcare services while improving the overall healthcare experience for patients and their families. The research will offer a scientific foundation for the future deployment of digital health solutions in LMICs, contributing to the development of evidence-based strategies that advance global child health, particularly in resource-constrained settings.

Methods

Data source

We used data from the service provision assessment (SPA), a nationally representative health facility survey implemented under the Demographic and Health Surveys (DHS) Program, funded by USAID. ²⁴ The SPA provides comprehensive assessments of health service availability and quality in LMICs through standardized methodologies, enabling cross-country comparisons. Each SPA survey includes a nationally representative sample of health facilities from both public and private sectors and across all levels of care—from tertiary hospitals to community health posts. Multiple instruments are used for data collection, including facility inventories, direct clinical observations, provider interviews, and client exit interviews. These methods allow simultaneous assessment of structural readiness (e.g., infrastructure, equipment, diagnostics, and medicines) and process quality (e.g., adherence to clinical guidelines, patient-centered communication). ²⁵

We selected countries that conducted SPA surveys involving sick child care observations and client exit interviews between 2014 and 2022. Where multiple datasets existed for a single country, only the most recent data were included. Our study comprises data from eight countries: Afghanistan (2019), the Democratic Republic of Congo (2018), Ethiopia (2022), Haiti (2018), Malawi (2014), Nepal (2021), Senegal (2019), and Tanzania (2015). In the pooled dataset, we derived key measures from multiple components of the SPA survey. The facility inventory module was used to construct the Digital Health Capacity Index, capturing the structural readiness of each facility. Clinical quality—operationalized as evidence-based practice—was assessed through direct observations of provider-client interactions during sick child consultations, where trained enumerators recorded provider adherence to standardized diagnostic and treatment protocols. Patient-centered care was measured using structured client exit interviews, which captured caregiver-reported experiences during the visit. Our study includes 5311 facilities and 20,880 sick children pooled for analysis: Afghanistan (84 facilities, 576 children), Democratic Republic of Congo (992, 2673), Ethiopia (813, 3755), Haiti (642, 2168), Malawi (748, 3441), Nepal (764, 2420), Senegal (253, 885), and Tanzania (1,015, 4962).

Measures and variables

Digital health capacity

We assessed digital health capacity at the facility level using the World Health Organization's Classification of Digital Health Interventions (CDHI), version 2.0.^8,26 The CDHI offers a globally recognized taxonomy for organizing and evaluating digital health functions, categorizing interventions by their primary users (e.g., clients, healthcare providers, system managers) and their intended functions (e.g., data services, communication, service delivery support). This classification provides a comprehensive and standardized framework to assess the structural readiness of health systems for digital transformation, particularly in low- and middle-income countries. ²⁷

Guided by this framework and WHO technical guidance, we constructed a composite Digital Health Capacity Index (DHCI) using five domains: (1) digital infrastructure, (2) clients, (3) healthcare providers, (4) data services, and (5) health system managers (see Supplementary eTable 1). Each domain was operationalized through binary indicators from the Service Provision Assessment survey. For domains composed of multiple items (e.g., digital infrastructure and data services), conservative scoring rules were applied to ensure minimum functionality. The five domain scores were summed to construct a facility-level index ranging from 0 to 5. For visualization purposes (e.g., heatmaps), the score was rescaled to a 0–1 range by dividing by 5. The total DHCI score reflects the overall digital readiness of a facility, while individual domain scores capture specific functional capacities.

Following formative measurement principles, ²⁸ we did not evaluate internal consistency, as each domain was conceptually distinct. As a robustness check, a principal component analysis (PCA) was conducted (Supplementary eFigure 1), and the PCA-derived score was highly correlated with the additive index (Pearson's r = 0.731, N = 20,342), supporting empirical validity.

Evidence-based practice

We assessed EBP using 10 binary clinical indicators derived from the WHO Integrated Management of Childhood Illness (IMCI) guidelines ²⁹ and validated through the Good Medical Practice Index (GMPI) developed by Lewis et al. ³⁰ using SPA data across LMICs. These indicators reflect core clinical activities (history-taking, physical examination, and caretaker counseling) required for appropriate diagnosis and management of sick children (see Supplementary eTable 2). All 10 items are routinely collected in SPA surveys and are aligned with international pediatric care standards. Validation studies have demonstrated high internal consistency, reliability, and construct validity, supporting their use as a robust indicator of clinical performance in LMICs.^30,31

To assess measurement robustness, we conducted a sensitivity analysis using an expanded 20-item quality tool developed by Macarayan et al., ³² which captures a broader range of clinical competencies. Results using this alternative definition of EBP were highly consistent with those from the 10-item index, supporting the stability and criterion validity of our primary measure.

Patient-centered care

PCC was measured using standardized patient-reported experience data from client exit interviews in the SPA survey. We initially selected 11 items spanning provider communication, responsiveness, respect for privacy, and facility conditions, each scored on a 3-point ordinal scale (1 = major problem, 2 = minor problem, 3 = no problem), with higher scores indicating better patient experience. Cronbach's alpha for the full 11-item scale was 0.79. To improve psychometric validity and reduce item redundancy, we conducted exploratory factor analysis (EFA) using principal axis factoring with oblique rotation. Seven items met the loading threshold (≥0.60) and formed a three-factor solution: communication, privacy/respect, and facility conditions (see Supplementary eTable 3). These dimensions aligned well with international frameworks including Consumer Assessment of Healthcare Providers and Systems (CAHPS^®), ³³ Primary Care Assessment Tool (PCAT), ³⁴ and WHO Framework on Integrated People-Centred Health Services (IPCHS), ³⁵ supporting theoretical and construct validity (see Supplementary eTable 4).

The refined 7-item PCC scale (range: 3–21) demonstrated strong internal consistency (Cronbach's alpha = 0.78) and was used as the primary measure (see Supplementary eTable 5). Criterion-related validity was assessed using patient satisfaction as an independent variable in multilevel regression models, which showed a strong positive association (β = 0.929, 95% CI: 0.864 to 0.993, p < 0.001; see Supplementary eFigure 2).

Covariates

We assessed factors influencing clinical quality at the individual, provider, and facility levels, with covariates selected based on established determinants of healthcare quality.^24,30,36 Individual-level variables included the child's gender (male or female), age, and the caretaker's relationship (mother or non-mother). Provider-level variables comprised provider type (physician, nurse, or other), years of experience, weekly working hours, and the level of support, defined as having at least two of the following: a written job description, promotion opportunities, or non-monetary incentives. Facility-level variables included managing authority (government or private), location (urban or rural), and facility type (hospital or clinic).

Statistical analyses

The study commenced with a descriptive analysis of the characteristics of clients, healthcare providers, and health facilities. For continuous variables, means and standard errors were calculated, while categorical variables were summarized using frequencies and percentages. To ensure national representativeness across the eight countries, all descriptive analyses were weighted based on facility, healthcare worker, and client samples. Additionally, a heatmap was generated to visualize the levels of digital health capacity and its sub-dimensions across countries. To facilitate visual comparison across countries and sub- dimensions, descriptive summaries and heatmaps rescaled the index to a standardized range from 0 to 1. This rescaling allows for intuitive cross-national interpretation of digital readiness levels while maintaining analytical consistency.

A multilevel linear regression model was employed to examine the relationship between digital health capacity and the implementation of Evidence-Based Practices (EBP) and Patient-Centered Care (PCC). We used the unweighted additive Digital Health Capacity Index (0–5) in regression models, assigning one point per domain. This preserved effect size interpretability and followed standard formative index methodology. Initially, the total digital health capacity score was used as the independent variable to assess its association with clinical quality, both with and without control variables. Subsequently, digital health was divided into five sub-dimensions, each treated as an independent variable, to further investigate their relationship with clinical quality. Sensitivity analyses were conducted to ensure the robustness of the results and to confirm that the findings were consistent under different model specifications. To control for extreme values, Winsorization was applied to both the PCC and EBP metrics, with extreme values replaced by those corresponding to the 1st and 99th percentiles. All regression analyses were performed without weighting. Statistical analyses were performed using Stata version 17.

Results

Table 1 shows the demographics and characteristics of facilities, health providers, caretakers, and children in the digital health study across eight countries. The study analyzed 20,699 observations from Afghanistan (2.77%, n = 574), the Democratic Republic of Congo (12.85%, n = 2660), Ethiopia (18.08%, n = 3742), Haiti (10.46%, n = 2166), Malawi (16.08%, n = 3329), Nepal (11.51%, n = 2383), Senegal (4.27%, n = 884), and Tanzania (23.97%, n = 4961). Out of 4705 surveyed facilities, 35.99% (n = 1693) were urban and 64.01% (n = 3012) were rural. Management-wise, 66.86% (n = 3146) were government-run, while 33.14% (n = 1559) were privately managed. Clinics were constituted 89.54% (n = 4213) of the facilities, with hospitals comprising 10.46% (n = 492). The dataset included 7360 healthcare providers: physicians (39.17%, n = 2883), nurses (38.75%, n = 2852), and others (22.08%, n = 1625). Providers averaged 6.24 years of experience (SD = 0.14), worked 50.84 h per week (SD = 0.58), and had 15.47 years of education (SD = 0.10). The support environment scored a mean of 0.34 (SD = 0.01). Among 20,694 caretakers, 81.94% (n = 16,957) were mothers and 18.06% (n = 3737) were non-mothers. The children's mean age was 20.38 months (SD = 0.18), with a gender distribution of 52.91% male (n = 10,951) and 47.09% female (n = 9748).

Table 1.

Characteristics of the facility, provider, caretaker, and sick child.

Characteristics	Weighted sample	Value
Country, No. (%)	20,699
Afghanistan		574	2.77
Congo Democratic Republic		2660	12.85
Ethiopia		3742	18.08
Haiti		2166	10.46
Malawi		3329	16.08
Nepal		2383	11.51
Senegal		884	4.27
Tanzania		4961	23.97
Facility
Location, No. (%)	4705
Urban		1693	35.99
Rural		3012	64.01
Manage authority, No. (%)	4705
Government		3146	66.86
Private		1559	33.14
Type, No. (%)	4705
Hospital		492	10.46
Clinic		4213	89.54
Provider
Type, No. (%)	7360
Physician		2883	39.17
Nurse		2852	38.75
Others		1625	22.08
Years of working, mean (SD)	7728	6.24	0.14
Working hours per week, mean (SD)	7728	50.84	0.58
Years of education, mean (SD)	7728	15.47	0.1
Support environment, mean (SD)	7728	0.34	0.01
Caretaker
Child relationship, No. (%)	20,694
Mother		16,957	81.94
Non-mother		3737	18.06
Child
Age of the month	19,950	20.38	0.18
Gender	20,699
Male		10,951	52.91
Female		9748	47.09

Note: Values are presented as n (%) for categorical variables and mean (standard deviation) for continuous variables. All statistics are weighted using the Service Provision Assessment (SPA) survey weights, which account for the complex sampling design across facilities, healthcare providers, and client levels to ensure national representativeness.

The heatmap is presented in Figure 1. The horizontal axis denotes the number of countries used in our study whereas, the vertical axis represents the total digital health score and the scores for each category. Red color indicates lower scores, while blue indicates higher scores. From the figure, it can be seen that the average digital health score across LMICs is 0.35. The scores for the five sub-dimensions are as follows: 0.02 for digital infrastructure, 0.52 for client engagement, 0.21 for healthcare providers, 0.92 for data services, and 0.06 for health system managers. The country with the highest total digital health score is Senegal, with a score of 0.39. Scores for digital infrastructure and health system managers are consistently low across all the countries however; the scores for clients and healthcare providers are moderate. The overall scores for data service are consistently high among each country, precisely: Afghanistan (0.59), Democratic Republic of the Congo (0.99), Ethiopia (0.69), Haiti (0.92), Malawi (0.89), Nepal (0.97), Senegal (0.95), and Tanzania (0.92).

Figure 1.

National variation in digital health capacity and sub-dimensions across eight LMICs. Note: Heatmap showing the mean facility-level scores for digital health capacity and its five functional domains across eight LMICs. Data are weighted and derived from SPA surveys (2014–2022). Digital Health scores are rescaled from 0–5 to 0–1 for visualization purposes. Original regression analyses were conducted using the unstandardized total score (range: 0–5). “Mean” represents the average across countries.

Table 2 shows the multilevel linear regression results assessing the relationship between digital health capacity and EBP. We have found positive and significant associations for total digital health (β=0.146, P < 0.001); digital infrastructure, 0.183 (P = 0.029); clients 0.205 (P < 0.001); and health providers 0.142 (P < 0.001). For covariates, the support environment also found a strong positive association with coefficients of 0.204 (P < 0.001). Conversely, the negative associations were identified for the age of the sick child −0.006 (P < 0.001), and non-mother caretakers −0.107 (P < 0.001). Other provider types also highlighted the negative association of −0.288 (P < 0.001), and years of working found a negative association of −0.007 (P = 0.005). Facility characteristics further influenced results, with rural facilities exhibiting positive associations of 0.168 (P < 0.001), whereas clinics, compared to hospitals, showed a negative impact of −0.104 (P = 0.049). Our results highlighted the complex impact of digital health components on clinic quality, highlighting the critical roles of digital infrastructure and supportive environments in enhancing healthcare outcomes.

Table 2.

Associations between digital health capacity and evidence-based practice: multilevel linear regression results.

Variable	M1 Coef.	95%CI	M1a Coef.	95%CI	M2 Coef.	95%CI	M2a Coef.	95%CI
Digital health (total) a	0.146***	[0.10,0.19]	0.111***	[0.07,0.16]
Digital infrastructurea					0.183*	[0.02,0.34]	0.13	[−0.04,0.30]
Clientsc					0.205***	[0.12,0.29]	0.173***	[0.09,0.26]
Health providersb					0.142***	[0.06,0.23]	0.122**	[0.04,0.21]
Data servicesa					0.026	[−0.13,0.18]	−0.057	[−0.21,0.10]
Health system managersa					0.111	[−0.02,0.24]	0.083	[−0.05,0.22]
Child
Gender (Ref. = Female)c			0.004	[−0.04,0.05]			0.007	[−0.03,0.05]
Age of sick child (month)c			−0.006***	[−0.01−−0.00]			−0.006***	[−0.01,−0.00]
Caretaker
Non-mother (Ref. = Mother)c			−0.107***	[−0.17,−0.05]			−0.104***	[−0.17,−0.04]
Age of caretaker (year)c			−0.002	[−0.01,0.00]			−0.002	[−0.01,0.00]
Provider
Nurse (Ref. = Physician)b			−0.03	[−0.13,0.07]			−0.014	[−0.11,0.09]
Others			−0.288***	[−0.40,−0.18]			−0.278***	[−0.39,−0.17]
Years of workingb			−0.007**	[−0.01,−0.00]			−0.008**	[−0.01,−0.00]
Working hours per dayb			0	[−0.00,0.00]			0	[−0.00,0.00]
Years of educationb			0.004	[−0.01,0.02]			0.005	[−0.01,0.02]
Support environmentb			0.204***	[0.14,0.27]			0.205***	[0.14,0.27]
Facility
Private (Ref. = Government)a			0.102*	[0.01,0.19]			0.084	[−0.01,0.17]
Rural (Ref. = Urban)a			0.168***	[0.08,0.25]			0.191***	[0.10,0.28]
Clinic (Ref. = Hospital)a			−0.104*	[−0.20,−0.00]			−0.117*	[−0.22,−0.01]

Note: Model (M) 1 and M1a examine the association between the total digital health score and EBP, with M1a adjusting for covariates. M2 and M2a assess individual digital health domains, with M2a including full covariate adjustment.

^arepresents facility level; ^brepresents healthcare provider level; ^crepresents client level.

* P < 0.05, ** P < 0.01, *** P < 0.001

Coefficient values such as 0.000 or −0.000 indicate very small but non-zero effects;

Confidence intervals such as [−0.01, 0.00] reflect small estimated effects near zero, rounded to two decimal places. These do not imply exact zero bounds.

Table 3 presents the results of the multilevel linear regression analysis examining the relationship between digital health capacity and PCC. The analysis showed that the regression coefficient of total digital health on PCC was negative but non-significant at −0.013 (P = 0.531). After controlling for confounding factors, the coefficient value was positive but still, it was non-significant at 0.044 (P = 0.098). Digital infrastructure also showed a negative relationship before and after controlling for confounding factors at −0.157 (P = 0.076) and −0.091 (P = 0.321), respectively. Clients showed a significant negative association of −0.136 (P = 0.006), but after controlling for confounding factors, a non-significant association of −0.08 (P = 0.119) was observed. Health providers have demonstrated positive and significant relationships before and after controlling for confounding factors with 0.122 (P = 0.023) and 0.115 (P = 0.034). Data services found a non-significant positive association before controlling for confounding factors, with a coefficient of 0.046 (P = 0.604), while controlling for confounding factors found a significant positive association of 0.210 (P = 0.023). Other provider types showed positive associations with 0.179 (P = 0.005), and years of education also had positive relationship of 0.021 (P = 0.005). The private facilities exhibiting positive associations of 0.552 (P < 0.001).

Table 3.

Associations between digital health capacity and patient-centered care: multilevel linear regression results.

Variable	M3 Coef.	95%CI	M3a Coef.	95%CI	M4 Coef.	95%CI	M4a Coef.	95%CI
Digital health (total) a	−0.013	[−0.06,0.04]	0.044	[−0.01,0.10]
Digital infrastructure a					−0.157	[−0.34,0.02]	−0.091	[−0.28,0.09]
Clients c					−0.136**	[−0.23,−0.04]	−0.08	[−0.18,0.02]
Health providers b					0.122*	[0.02,0.22]	0.115*	[0.01,0.22]
Data services a					0.046	[−0.13,0.22]	0.210*	[0.03,0.39]
Health system managers a					0.107	[−0.04,0.25]	0.117	[−0.04,0.27]
Child c
Gender (Ref. = Female) c			0.008	[−0.05,0.06]			0.007	[−0.05,0.06]
Age of sick child (month) c			−0.002*	[−0.00,−0.00]			−0.002	[−0.00,0.00]
Caretaker c
Non-mother (Ref. = Mother) c			−0.038	[−0.12,0.04]			−0.047	[−0.13,0.03]
Age of caretaker (year) c			0.001	[−0.00,0.00]			0	[−0.00,0.00]
Provider b
Nurse (Ref. = Physician) b			−0.083	[−0.20,0.03]			−0.103	[−0.22,0.02]
Others			0.179**	[0.05,0.31]			0.150*	[0.02,0.28]
Years of working b			0.005	[−0.00,0.01]			0.005	[−0.00,0.01]
Working hours per day b			0	[−0.00,0.00]			0	[−0.00,0.00]
Years of education b			0.021**	[0.01,0.04]			0.019*	[0.00,0.03]
Support environment b			0.044	[−0.04,0.13]			0.044	[−0.04,0.13]
Facility a
Private (Ref. = Government) a			0.552***	[0.45,0.65]			0.579***	[0.48,0.68]
Rural (Ref. = Urban) a			0.073	[−0.02,0.17]			0.037	[−0.07,0.14]
Clinic (Ref. = Hospital) a			0.347***	[0.23,0.46]			0.345***	[0.23,0.46]

Note: Model (M)3 and M3a examine the association between the total digital health score and PCC, with M3a adjusting for covariates. M4 and M4a assess individual digital health domains, with M4a including full covariate adjustment.

^arepresents facility level; ^b represents healthcare provider level; ^c represents client level.

* P < 0.05, ** P < 0.01, *** P < 0.001

Coefficient values such as 0.000 or −0.000 indicate very small but non-zero effects;

Confidence intervals such as [−0.00, 0.00] reflect small estimated effects near zero, rounded to two decimal places. These do not imply exact zero bounds.

Table 4 illustrates the results of the sensitivity analysis between digital health capacity, EBP, and PCC (see eTables 6–9 in Supplement for details). For the EBP, total digital health demonstrated positive and significant connotations before and after controlling for confounding factors at 0.249 (P < 0.001) and 0.184 (P < 0.001) respectively. The winsorized model found the relationship at 0.145 (P < 0.001) and 0.110 (P < 0.001). Healthcare providers and clients also demonstrated a significant positive relationship before and after controlling for confounding factors.

Table 4.

Sensitivity analyses of the associations between digital health and EBP and PCC outcomes.

Variable	M5 Coef.	95%CI	M5a Coef.	95%CI	M6 Coef.	95%CI	M6a Coef.	95%CI
EBP-20 items
Digital health	0.249***	[0.18,0.32]	0.184***	[0.11,0.26]
Digital infrastructure					0.192	[−0.06,0.45]	0.091	[−0.17,0.35]
Clients					0.399***	[0.27,0.53]	0.325***	[0.19,0.46]
Healthcare providers					0.299***	[0.17,0.42]	0.241***	[0.11,0.37]
Data services					0.227	[−0.01,0.47]	0.056	[−0.19,0.30]
Health system managers					0.003	[−0.20,0.21]	−0.017	[−0.23,0.20]
EBP 1st-99th winsorization
Digital health (total)	0.145***	[0.10,0.19]	0.110***	[0.06,0.16]
Digital infrastructure					0.182*	[0.02,0.34]	0.128	[−0.04,0.29]
Clients					0.203***	[0.12,0.29]	0.171***	[0.09,0.26]
Healthcare providers					0.140***	[0.06,0.22]	0.119**	[0.03,0.20]
Data services					0.028	[−0.12,0.18]	−0.055	[−0.21,0.10]
Health system managers					0.11	[−0.02,0.24]	0.081	[−0.05,0.22]
PCC-11 items
Digital health (total)	−0.108***	[−0.19, −0.03]	0.039	[−0.04, 0.12]
Digital infrastructure					−0.456**	[−0.74, −0.17]	−0.240	[−0.53, 0.05]
Clients					−0.306***	[−0.46, −0.15]	−0.162*	[−0.32, −0.01]
Healthcare providers					0.173*	[0.01, 0.33]	0.209*	[0.05, 0.37]
Data services					−0.150	[−0.43, 0.13]	0.152	[−0.13, 0.43]
Health system managers					0.122	[−0.11, 0.36]	0.162	[−0.08, 0.40]
PCC 1st-99th winsorization
Digital health (total)	−0.012	[−0.06,0.04]	0.043	[−0.01,0.09]
Digital infrastructure					−0.152	[−0.33,0.02]	−0.088	[−0.27,0.09]
Clients					−0.133**	[−0.23,−0.04]	−0.078	[−0.17,0.02]
Healthcare providers					0.124*	[0.03,0.22]	0.117*	[0.02,0.22]
Data services					0.029	[−0.14,0.20]	0.187*	[0.01,0.36]
Health system managers					0.103	[−0.04,0.25]	0.111	[−0.04,0.26]

Note: Model (M) 5 evaluates the association between the total Digital Health Capacity Index and each outcome: (i) EBP measured by 20 items; (ii) EBP with winsorized values; (iii) PCC measured using a validated 11-item score; (iv) PCC with winsorized values, and Model 5a includes covariates. M6 assesses each digital health domain, and M6a includes covariates.

* P < 0.05, ** P < 0.01, *** P < 0.001

For PCC, results remained consistent across different specifications. In the PCC-11 items model, total digital health showed a significant negative association before covariate adjustment (−0.108, P = 0.006), which became non-significant after adjustment (0.039, P = 0.300). Digital infrastructure and clients also demonstrated negative relationships in both models, with digital infrastructure showing a significant association (−0.456, P = 0.002) before adjustment. In contrast, healthcare providers were positively associated with PCC, with coefficients of 0.173 (P = 0.036) and 0.209 (P = 0.012) before and after controlling for covariates, respectively. Similarly, in the winsorized PCC model, total digital health had a non-significant connection in both models −0.012 (P = 0.598) and 0.043 (P = 0.101) correspondingly. Digital infrastructure and clients also highlighted the negative associations in both models, with one significant association for clients −0.133 (P = 0.006). Health providers have positive and significant links 0.124 (P = 0.020) and 0.117 (P = 0.022) before and after controlling for confounding factors.

Additionally, Supplement eTable 10 presents sensitivity analyses adjusting for pediatric consultation volume. The results confirm that the association between digital health and EBP remained significant and robust (Coef. = 0.116, P < 0.001) after accounting for facility-level variation in child caseload. Associations with PCC and individual digital health domains were also consistent in magnitude and direction, supporting the stability of our main findings across different facility types.

Discussion

Our study characterizes the foundational landscape of digital health technology in LMICs and examines its influence on influence on evidence-based practice and patient-centered care. Drawing on validated, multidimensional constructs of digital health capacity, EBP, and PCC, we found that digital health capacity—though underdeveloped in many LMICs—was significantly associated with improvements in clinical performance (EBP), but had limited influence on patient experience (PCC).

These results align with previous studies. For example, Mansah et al., reported positive outcomes in seven of nine dimensions related to consumer experiences post-digitalization. ³⁷ Similarly, Ndayishimiye et al.'s comprehensive review during the COVID-19 pandemic revealed that digital health tools were pivotal in facilitating virtual healthcare, providing clinical assistance, monitoring care quality, and managing critical health resources.^38,39 Unlike most prior research focused on chronic care in high-income countries with robust ICT infrastructure (e.g., Australia, Canada, the UK, and the US),^17,18 our study centers on pediatric care in LMICs and uses harmonized measures to enable cross-country comparison.

Our domain-specific analysis offers additional insight. Digital infrastructure was positively associated with EBP, reinforcing the importance of stable internet connectivity, reliable electricity, and computing equipment for digital interventions.^39–41 These results echo prior findings from China and elsewhere emphasizing the foundational role of digital infrastructure in expanding access to care and enabling advanced tools like telemedicine and electronic health records. ⁴²

In terms of client interactions, —such as patient feedback systems—also showed strong positive associations with EBP, suggesting that systems which incorporate patient voice can support higher quality clinical encounters. These results are consistent with literature underscoring the value of feedback loops in quality improvement and patient-centered systems.^43,44 Studies, such as those by Rachel et al., demonstrate that incorporating patient perspectives significantly enhances patient experiences and identifies areas for service improvement. ⁴⁵ Hiyam et al. propose that to enhance service quality, control systems should prioritize the dissemination phase of patient survey results among healthcare professionals. ⁴⁶ However, we also found a negative association between client feedback tools and PCC, indicating that implementation of digital tools may not automatically translate into a better patient experience. Possible explanations include superficial adoption, limited patient engagement, or increased administrative burden on providers. ⁴¹

Among healthcare providers, training and digital support systems were robustly associated with improved EBP.^47,48 These findings affirm the importance of equipping providers with the tools and competencies to navigate digital systems and leverage health information for clinical decision-making. In line with existing literature, facilities with trained providers using health management information systems (HMIS) were better positioned to deliver guideline-based care. ⁴⁹

The associations between data systems and EBP were weaker and inconsistent, highlighting potential challenges in the quality or usage of digital data systems in LMICs. While routine data reporting is essential for performance monitoring, its direct clinical utility may depend on real-time accessibility, provider capacity, and integration with clinical workflows. ⁵⁰ Similarly, digital capacity among health system managers—such as inventory tracking and supervision—showed limited direct associations with care outcomes, possibly reflecting the structural rather than frontline nature of these functions. Studies have demonstrated that effective inventory management using digital health tools significantly improves operational efficiency and healthcare outcomes by ensuring timely access to medications and optimizing the use of financial and human resources. ⁵¹ However, in low-income countries, the absence of robust data collection and reporting systems, coupled with limited adoption of digital health tools like inventory management systems, hampers their impact on healthcare quality.^52,53

Notably, total digital health capacity was not significantly associated with patient-centered care (PCC) across most model specifications. This finding is consistent with prior literature suggesting that digitalization alone may not translate into improved patient-reported experiences, particularly in resource-constrained settings characterized by low health literacy, limited patient engagement, and contextual factors such as long wait times or poor facility conditions. ⁵⁴ One plausible explanation is that PCC is inherently shaped by patient expectations, which may be modest in low- and middle-income country contexts. Consequently, a positive user experience may be reported despite limited digital integration, and conversely, higher digital capacity does not guarantee improved perception if these contextual factors are not addressed. Moreover, environmental elements such as cleanliness, noise levels, and the availability of basic amenities (e.g., food provision) have been shown to significantly influence user experience in hospital settings. ⁵⁵ Core components of optimized PCC include timely access to care, patient engagement, a strong patient-physician relationship, evidence-based care, clear communication, physical and emotional support, family involvement, personalized care, service responsiveness, and continuity of care.^56–58 While digital health tools may enhance some of these domains—such as communication or follow-up tracking—their effectiveness likely depends on how they are integrated with broader care delivery systems and whether they are accompanied by institutional efforts to enhance responsiveness and human-centered service design.

Despite the significant findings, our study has several limitations. First, the cross-sectional design restricts our ability to draw causal inferences regarding the observed associations. Second, although the digital health, EBP, and PCC measures were rigorously developed using validated indicators and frameworks, they were inherently limited by the scope of available SPA data and may not capture the full complexity of digital functionality or patient experience. Third, while the dataset covers eight LMICs, findings may not be generalizable to countries with substantially different health system structures or levels of digital health maturity. Finally, despite comprehensive sensitivity analyses, future research should incorporate longitudinal designs and mixed-method approaches to better understand underlying mechanisms and long-term effects.

Conclusion

Our study demonstrates that digital health capacity significantly improve EBP in LMICs, particularly in relation to digital infrastructure, patient (feedback systems), and provider (training). Its influence on patient experience is more nuanced and may require complementary strategies to translate digital investment into perceived care improvements. These findings underscore the need for comprehensive approaches that address digital infrastructure, training, and patient engagement to optimize the benefits of digital health in enhancing pediatric healthcare services in resource-limited settings.