Healthcare providers’ knowledge, attitudes, and practices in nutrition management for patients with type 2 diabetes mellitus in Kinshasa: a cross-sectional study in Democratic Republic of the Congo

Abstract

Background

Nutrition is critical for diabetes management, yet evidence about healthcare providers’ knowledge, attitudes, and practices (KAP) in sub-Saharan Africa, including the Democratic Republic of the Congo (DRC), remains limited. We evaluated healthcare providers’ KAP related to nutrition management for patients with type 2 diabetes in Kinshasa and explored links between knowledge, attitudes, and practice.

Methods

We conducted a descriptive correlational cross-sectional study between November and December 2024 among healthcare providers in diabetes care in Kinshasa, DRC. Using probabilistic multi-stage sampling of healthcare facilities across all 35 health zones, 877 healthcare providers were interviewed. KAP scales were scored across sociodemographic and professional characteristics and family history of diabetes. Structural equation modelling estimated direct and indirect effects among knowledge, attitude, and practice and evaluated determinants.

Results

Among 877 healthcare providers, there were 267 doctors, 574 nurses, and 36 nutritionists. The median age was 39 years (IQR = 32–46), and a minority had received diabetes-related training within the past 12 months (232/877; 26%). Mean scores for knowledge were 11.9/19 (SD = 2.6), for attitude 11.9/14 (SD = 2.9), and for practice 10.0/17 (SD = 5.4). While most healthcare providers (809/877; 89%) agreed that nutrition is fundamental for diabetes care, many healthcare providers viewed nutritional management for hospitalised patients as the nutritionist’s sole responsibility. Many healthcare providers reported that they did not utilise any record form for nutritional assessment (499/877; 56.9%) or any record for nutrition-related issues or diagnoses (446/877; 50.9%). Structural equation modelling showed that both knowledge (β = 0.54, p < 0.001) and attitudes’ scores (β = 0.81, p < 0.001) significantly influenced practices, with attitude mediating the effect of knowledge on practice. Professional category and training were the strongest predictors of higher KAP scores, whereas female sex was negatively associated with scores for both knowledge and attitude.

Conclusion

The study revealed major gaps in healthcare providers’ knowledge and practices on nutrition management in Kinshasa, shaped by both knowledge and attitudes. Differences by profession and sex highlight training inequities. Despite its cross-sectional and self-reported limitations, the study highlights the need for targeted, context-specific interventions to strengthen frontline healthcare providers’ skills and attitudes on that aspect in urban sub-Saharan Africa.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12913-026-14799-2.

Introduction

Diabetes mellitus (DM) refers to metabolic disorders marked by chronic hyperglycaemia, caused by impaired insulin secretion, insulin action, or both [1]. In 2024, diabetes affected 589 million adults (20–79 years) and caused over 3.4 million deaths, representing 9.3% of global mortality. About 90% of cases are type 2 diabetes (T2DM) [2]. Once considered a disease of wealthy nations, T2DM now shows the fastest growth in low- and middle-income countries (LMICs), especially in sub-Saharan Africa [3]. In Africa, 24.6 million adults lived with diabetes in 2024 (5% prevalence), a figure projected to rise by 142% to 59.5 million by 2050 [2]. Poorly managed T2DM substantially adds to the burden of chronic diseases by increasing the risk of complications such as stroke, cardiovascular disease, kidney failure, poor circulation and neuropathy leading to lower limb amputation, and diabetic retinopathy leading to blindness [4]. Effective management relies on a combination of pharmacological treatment and lifestyle modification, including diet modification and good sleeping quality [5–7].

Medical nutritional therapy refers to the planned use of dietary assessment, individualised dietary advice, nutrition education, and follow-up to support glycaemic control, weight management, cardiovascular risk reduction, and overall well-being in people living with diabetes. It is a cornerstone of type 2 diabetes mellitus management and may include strategies such as improving overall diet quality, adapting carbohydrate intake and distribution, increasing dietary fibre intake, limiting saturated fat intake, and supporting sustainable behaviour change [6, 8, 9, 10]. Nutrition education and regular follow-up are essential to ensure patient adherence, improve glycaemic control, and reduce complications [11–13]. The latest American Diabetes Association (ADA) standard of care reiterated the importance of medical nutrition therapy and counselling for behaviour change in improving T2DM control and patient well-being [8]. In this study, counselling for behaviour change refers to structured provider-patient communication that supports patients to adopt and maintain behaviours such as healthy eating, physical activity, medication adherence, smoking cessation, and reduced alcohol consumption [8, 14, 15].

However, in sub-Saharan Africa, optimal T2DM management is limited by shortages of trained healthcare providers, weak infrastructure, lack of diagnostic tools, and high treatment costs. Sociocultural factors, such as communal eating practices; the perception that being overweight is healthy and associated with wealth and prestige; lack of resources to buy healthy foods; and poor disease awareness, are all important challenges for African T2DM patients [16–18] and are highly prevalent in the DRC [18–21]. Recent studies on service delivery for diabetes management in the DRC have found a range of challenges, including inadequate staffing and insufficient essential drugs [22–24].

Healthcare providers play a critical role in T2DM prevention and management, particularly in primary and secondary healthcare settings where they are often the first point of contact. Their knowledge, attitudes, and practices (KAP) regarding nutrition are key determinants of patient outcomes [25, 26]. The ADA standard of care explicitly identified clinicians’ counselling skills as one of the factors that can drive adherence to nutrition plans, thereby improving diabetic control. Counselling skills mean the ability to give dietary/lifestyle advice, assess patient needs, use patient-centered communication, negotiate goals, and offer follow-up support [6, 8, 25]. Inadequate healthcare provider’s knowledge or suboptimal counselling skills can contribute to poor glycaemic control and increased complication rates [27]. Understanding the pathways linking healthcare providers’ KAP is essential for strengthening their capacity in nutrition management for patients with T2DM [26]. Interventions to improve knowledge may not lead to better dietary counselling or patient support unless attitudes that drive consistent practice are also addressed [28].

While studies in Africa have examined patients’ perceptions [29] or provider KAP in other countries such as Senegal [30] and Mali [31], there is a lack of evidence on the healthcare providers’ KAP regarding nutrition management for patients with T2DM in the DRC. The few KAP studies conducted in the DRC have revealed gaps among healthcare providers’ knowledge and attitudes about diabetes management [22, 32, 33], but none have focused on the nutrition aspects of diabetic care, nor did they attempt to establish a link between knowledge, attitudes, and practices. To address this gap, this study assessed healthcare providers’ knowledge, attitudes, and practices regarding nutrition management for patients with T2DM in Kinshasa and examined the pathways linking these dimensions. The analysis was guided by a KAP framework in which knowledge is expected to influence practice both directly and indirectly through attitudes, consistent with previous pathway-based studies of diabetes-related knowledge, attitudes, and practices [26, 34–36].

Methods

Study design

This study used an analytical cross-sectional design with descriptive and correlational components. The descriptive component estimated the levels of knowledge, attitudes, and reported practices regarding nutrition management for patients with diabetes among healthcare providers in Kinshasa. The analytical component examined associations between provider/facility characteristics and KAP scores, as well as the pathways linking knowledge, attitudes, and practices. This design was appropriate because the study aimed to assess the distribution of KAP and their interrelationships at a single point in time [37, 38]. However, because exposure and outcome variables were measured simultaneously, the study cannot establish temporality or causality. This is an in-depth analysis of data from a larger study that explored healthcare providers’ KAP on diabetes mellitus and hypertension (unpublished report). The reporting of this study was guided by the STROBE recommendations for observational cross-sectional studies [39].

Study setting

This study took place in Kinshasa, the capital and most urbanised city of the DRC, which, according to the 2024 Demographic and Health Survey, has the country’s highest diabetes prevalence among both women of reproductive age (9.5%) and men aged 15–59 (7.9%) [40]. Diabetes prevention and management in Kinshasa is provided through primary, secondary, and tertiary care. At the primary level, diabetic care is mainly undertaken by nurses, with oversight provided by doctors. In secondary facilities, diabetic care is usually provided by doctors, with the assistance of nurses, and nutritionists may also be available. Despite the paucity of nutritionists, there are no nutritional guidelines in place. Each healthcare provider draws upon his own personal and professional background, as well as the knowledge acquired through various sources. The staff of public and faith-based health facilities more easily benefit from the training organised occasionally by the Ministry of Health and its partners than those in the private sector. Faith-based facilities generally receive more patients, and therefore the medical staff (usually in limited numbers) are more in demand there. Diabetes care in the DRC is predominantly financed through out-of-pocket payments, as no national health insurance scheme currently provides comprehensive coverage for diabetes management. Although some community-based health insurance initiatives, such as the Mutuelle de Santé pour les Enseignants du Secteur Public (MESP), exist, their coverage remains limited. In all health facilities, there is no standardized pricing, creating a significant barrier to equitable access to diabetes services.

Participants

The study population consisted of medical doctors, nurses, and nutritionists working in public, faith-based, and private health facilities, including those affiliated with the army and police. Eligible participants included healthcare providers working in internal medicine or emergency departments in secondary healthcare facilities, as well as healthcare providers (physicians, nurses, or nutritionists) practising in primary healthcare facilities who had been involved in the care of patients with diabetes for at least six months. In clinical practice, doctors, nurses, and nutritionists all have complementary roles. Doctors should integrate nutrition into the overall diabetes care plan and identify patients requiring specialised dietary support; nurses should provide routine counselling, reinforce dietary advice, and monitor adherence during follow-up; and nutritionists should conduct more detailed nutrition assessment, individualised dietary planning, and follow-up when available [8, 26, 25, 41]. This study excluded tertiary facilities where diabetic care is provided only for complicated cases, not for routine care.

Sampling

The sample size was calculated, in the primary study, in order to determine KAP on T2DM and hypertension among healthcare providers using the formula for estimated a proportion, n ≥ deff*z² pq/[d²(1-c)]. Because no prior estimate of hypertension-related indicators among healthcare providers was available, the expected proportion (p) was conservatively set at 50%, which maximizes sample size. A 95% confidence level was assumed (Z = 1.96), with a margin of error (d) of 5%. A design effect (deff) of 2 was applied to account for clustering in the sampling design, and a non-response rate of 10% (c = 0.10) was anticipated. Based on these assumptions, the minimum required sample size was estimated at 854 healthcare providers.

The final sample included 880 healthcare providers, selected using a four-stage sampling method (Fig. 1): (1) Inclusion of all 35 health zones (HZ); (2) Simple random selection of two health areas (HA) per HZ; (3) Simple random selection of four healthcare facilities per HA stratified according to their authority management (4). Simple random selection of providers managing diabetic patients stratified according to professional category (nurses, doctors and nutritionists). Within each health zone, the general referral hospital (GRH) was also included, therefore 35 GRHs were selected. Within each health area, the four healthcare facilities (HFs) included one public HF, one faith-based HF, and two private HFs. However, if all categories of health facility were not present in the HZ, fewer health facilities were selected. As a result, we had an average of seven to eight HFs per HZ. In secondary HFs, healthcare providers were randomly selected and included two doctors, and two nurses, one each from internal medicine and emergency department, and one nutritionist per facility. In primary HFs, two nurses managing diabetic patients and one doctor were selected where they were presents.

Fig. 1.

Sampling strategy flow chart

Data collection

Data were collected between November and December 2024. Five data collection teams, each comprising three interviewers and one field supervisor, were trained in survey methodology and Kobo Collect before conducting a pretest in two non-selected health areas. The data collection tool was revised based on feedback from that pretest. Each team covered seven health zones in 18 days, with an additional day for completing any missing data. Each day, after selecting the healthcare providers to be interviewed, the fieldworkers conducted face-to-face interviews, completing the questionnaire section by section with each participant. This approach was selected to reduce missing data, ensure that all participants understood the response options, standardise administration across professional categories, and accommodate variable provider availability during working hours [42–45]. To minimise interviewer bias, all interviewers received standardised training, the tool was pretested, and supervisors reviewed data quality daily, consistent with good practice in structured survey administration and observational study reporting [39].

The principal investigator (SB) provided oversight in the field throughout the data collection process. Each day’s data was uploaded to the server, allowing him to continuously evaluate the quality and integrity of the information and offer timely feedback to the fieldworkers.

Variables and measures

Conceptual and operational definitions

Healthcare providers include professionals who deliver medical services to patients, such as physicians, nurses, and nutritionists [46]. For the purposes of this study, healthcare providers were defined as doctors, nurses, and nutritionists working in selected health facilities and involved in the care of patients with diabetes for at least six months.

Nutrition management for patients with diabetes was defined as the set of provider activities aimed at assessing nutrition-related needs, providing dietary and lifestyle counselling, documenting nutrition-related problems, referring to or coordinating with nutrition professionals when needed, and supporting follow-up of patients with diabetes. This definition is consistent with diabetes care recommendations emphasizing nutrition therapy, behaviour change support, and multidisciplinary care [8, 14].

Knowledge referred to providers’ factual understanding of recommended nutrition care for patients with diabetes, including macronutrients, dietary quality, lipid-related concepts, fruit and fibre intake, and lifestyle recommendations. Operationally, knowledge was measured using 19 true/false/don’t know items, scored 1 for a correct answer and 0 for an incorrect or “don’t know” answer, consistent with previous KAP and diabetes nutrition knowledge assessments [26, 47, 48].

Attitudes referred to providers’ beliefs, perceptions, and level of agreement with recommended roles and principles of nutrition management for patients with diabetes. Operationally, attitudes were measured using 14 Likert-scale statements, recoded as positive or non-positive according to the expected response agreed upon by stakeholders and aligned with diabetes care recommendations [8, 26].

Practices referred to self-reported provider behaviours related to dietary counselling, nutritional assessment, documentation, referral, coordination with nutritionists, and lifestyle advice for patients with diabetes. Operationally, practices were measured using 17 frequency-based items and recoded as adequate or inadequate according to expected standards for nutrition-related diabetes care [8, 14, 25, 26].

Structured questionnaires were administered on tablet computers using the Kobo Collect application. Two validated instruments, the Nutritional Management of Diabetes Knowledge Test [26, 47] and the Revised Diabetes Knowledge Scale, were adapted in accordance with current diabetes care guidelines [48]. The research protocol and data collection tools were reviewed collaboratively with key stakeholders, including the investigator team, the National Program for Diabetes Control, the National Program for Cardiovascular Diseases (both under the Ministry of Health), and the NGO “Santé Rurale” (SANRU) (for details, see supplementary Additional files 1 A-C).

The final knowledge section comprised 19 statements (Additional file 1 A), to which participants responded with “true,” “false,” or “don’t know.” The attitude section included 14 statements (Additional file 1B), rated on a five-point Likert scale: “strongly agree,” “somewhat agree,” “neutral,” “somewhat disagree,” and “strongly disagree.” The practices section consisted of 17 statements (Additional file 1 C), where respondents indicated the frequency of engagement using the options “never,” “sometimes,” “often,” or “always”. Each statement was read aloud, and participants were asked to indicate how strongly they agreed with the statement or how frequently they engaged in the described behavior.

Data for all KAP variables were obtained from the structured interviewer-administered questionnaire. For each scale, item scores were summed to create a total score. The knowledge score ranged from 0 to 19, the attitude score from 0 to 14, and the practice score from 0 to 17. Higher scores indicated better knowledge, more positive attitudes, and more adequate reported practices, respectively. Scores were analysed as continuous variables. Item-level proportions were also reported to identify specific knowledge, attitude, and practice gaps, as recommended for KAP studies and instrument-based assessments [26, 47, 48].

Reliability assessment

Internal consistency of the knowledge scale was assessed using the Kuder–Richardson formula 20 (KR-20), which is appropriate for dichotomously scored items. Knowledge items were first coded as 1 for correct and 0 for incorrect or “don’t know.” For each item, the proportion of correct responses and incorrect responses was calculated. The variance of the total knowledge score was then estimated, and KR-20 was computed using the standard formula:

where k is the number of items, p is the proportion of correct responses for each item, q is 1 − p, and σ² is the variance of the total knowledge score. KR-20 was selected because the knowledge scale was composed of dichotomous items and has been widely used to assess internal consistency for such scales [49]. The detailed calculation, item-level values, and reliability results are provided in Additional file 2.

Item-level diagnostics (item difficulty, corrected item-total correlations) were performed for the KAP scale. It refers to the statistical analyses conducted on individual items of a scale or questionnaire to assess their performance and contribution to the overall reliability of the instrument. These diagnostics help identify items that may not function as intended, allowing researchers to refine the measurement tool. Item difficulty was defined as the proportion of participants who answered each item correctly; values close to 0 indicated difficult items, while values close to 1 indicated easy items. Corrected item-total correlation assessed the correlation between each item and the total score calculated from the remaining items. Low corrected item-total correlations suggested that an item contributed weakly to the overall scale or measured a different subdomain. These diagnostics were used to interpret internal consistency and identify items requiring revision in future applications of the tool [49, 50].

Data analysis

Data analysis proceeded in five steps using Stata SE 18. First, raw questionnaire responses were checked for completeness and consistency. Second, items were recoded according to expected answers agreed upon by the investigator team and stakeholders and aligned with diabetes care recommendations. Third, item scores were summed to generate total knowledge, attitude, and practice scores. Fourth, descriptive and bivariate analyses were conducted to compare KAP scores across provider and facility characteristics. Fifth, structural equation modelling was used to examine the pathways linking knowledge, attitudes, and practices.

For the assessment of knowledge, attitudes and practices, a scoring system had been developed after consultation with all stakeholders and in alignment with national guidelines. For knowledge items, correct answers were coded 1, while incorrect and “don’t know” responses were coded 0. For attitude items, “strongly agree” and “somewhat agree” were grouped as agreement, while “somewhat disagree” and “strongly disagree” were grouped as disagreement. Depending on the expected direction of each statement, responses aligned with recommended diabetes nutrition care were coded 1 and responses not aligned were coded 0. Neutral responses were coded as 0 because they did not indicate clear agreement with the expected professional position. This grouping simplifies the analysis, allowing for a binary outcome (Agree vs. Disagree) that facilitates statistical tests. Combining responses into fewer categories helps to increase statistical power and interpretability. For practice items, responses were coded as adequate when the reported frequency matched the expected standard and inadequate when it did not. This approach was used to produce interpretable KAP scores and to identify clear alignment or non-alignment with recommended diabetes nutrition care practices [8, 26, 47, 48].

Continuous variables, such as age and years of working experience, were summarised using medians and interquartile ranges, while categorical variables were presented as absolute and relative frequencies. Bivariate analyses, including t-tests, were conducted to explore significant associations between dietary knowledge scores, dietary attitude scores, dietary practice scores, and categorical variables (e.g., sex and recent training status). Analysis of variance was conducted to explore significant associations between dietary knowledge scores, dietary attitude scores, dietary practice scores and facility type. Correlation analyses were conducted to explore correlations between dietary knowledge scores, dietary attitude scores, dietary practice scores and experience (in years).

This study utilised a structural equation modelling (SEM) framework to analyse the relationships between dietary knowledge, attitudes, and practices among 877 healthcare providers. Structural equation modelling is a multivariable analytical framework that estimates several regression equations simultaneously. It is useful when the research question involves pathways in which one variable may act both as an outcome and as a predictor of another variable. In this study, SEM was used because we hypothesised that knowledge would influence practice both directly and indirectly through attitudes, consistent with previous KAP pathway models in diabetes-related research [34–36]. A standard multiple linear regression model could estimate predictors of practice but would not simultaneously estimate the knowledge-to-attitude pathway, the attitude-to-practice pathway, and the indirect effect of knowledge on practice through attitudes. Therefore, SEM was better aligned with the conceptual KAP framework guiding this analysis [51–53]. The SEM analysis was conducted using maximum likelihood estimation, accounting for potential clustering effects by applying robust standard errors adjusted for clusters based on facility ID. The model specified pathways from exogenous variables (independent variables such sex, year of experience, professional categories, knowledge) to the endogenous variables (dependent variables, attitude and practice variables), enabling the assessment of direct, indirect, and total effects. In practical terms, a pathway represented an assumed directional relationship between variables in the KAP framework. For example, the pathway from knowledge to attitudes tested whether providers with higher knowledge scores also had more positive attitudes. The pathway from attitudes to practices tested whether providers with more positive attitudes reported better nutrition care practices. The indirect pathway from knowledge to practice through attitudes tested whether part of the association between knowledge and practice operated through attitudes. Similar pathway-based approaches have been used in KAP studies examining diabetes-related behaviours and self-management [34–36, 51]. In summary, the model evaluated how dietary knowledge and attitude scores mediated the relationship between the independent variables and dietary practice scores. Only variables that were significant in the bivariate analysis with dependent variables were included as covariates of each dependent variable in the SEM. The rationale for limiting covariates to those that were significant in the bivariate analysis is grounded in established statistical practices. This method aligns with the principles of parsimony in modeling, where simpler models that adequately explain the data are preferred over more complex models that may not add additional explanatory power. Furthermore, it allows for a more straightforward interpretation of the results, as each included variable has a demonstrated relationship with the dependent variables, thereby enhancing the clarity of the findings. By following this methodology, we aimed to construct a robust SEM that accurately represents the relationships among the variables and provides meaningful insights into the factors influencing nutrition management practices. The goodness of fit for the SEM was evaluated using the Standardised Root Mean Squared Residual (SRMR) and the Akaike Information Criterion (AIC) (see details for SEM in additional file 7). A satisfactory fit was indicated by an SRMR value < 0.08 and an AIC value indicating a lower value compared to competing models. Variance Inflation Factors (VIF) were calculated to assess multicollinearity among the independent variables. A mean VIF value of 1.15 and individual VIFs below 2 indicated that multicollinearity was not an issue in the analysis.

Results

General characteristics of healthcare providers

Table 1 describes the general characteristics of healthcare providers and their involvement in basic and continuing training on DM. In total, 254 HFs were visited, of which the majority (67%, 170/254) were at the primary level and the rest at the secondary level; many (68.9%, 175/254) were private HFs, while few were public (20.5%, 52/254) and faith-based HFs (10.6%, 27/254). A total of 877 healthcare providers agreed to participate among the 880 approached in the study, comprising predominantly men (55%, 483/877). Half of the participants were under 39 years old (IQR:32–46). Nurses comprised two-thirds (65%, 574/877) of the sample, while nutritionists accounted for only 4% (36/877). Slightly more than a quarter (26%, 232/877) had participated in continuing education on diabetes mellitus in the past 12 months. The distribution of participants by facility type (depending on its authority management) indicated that 35% (307/877) worked in public facilities, 49% (430/877) in private facilities, and 16% (140/877) in other settings.

Table 1.

Socio-economic, demographic characteristics of healthcare providers

Characteristics	n = 877	%
Sex
Men	483	55,1
Women	394	44.9
Education level
Secondary school (A2, A3)	122	13.9
Bachelor’s degree	408	46.5
Five-years higher education degree (A0, Diploma of doctor)	261	29.7
Professional category
Nurse	574	65.5
Medical Doctor	267	30.4
Nutritionist	36	4.1
Family member living with diabetes (yes)	327	37.3
Authority management of HF where the HCP was working
Faith-based	140	16.0
Public	307	35.0
Private	430	49.0
Participation in job training on diabetes the past 12 months	232	26.4
Year of experience in diabetes management [median (IQR)]	7years (4–12)

Secondary school for nurse refers to the high-school leaving diploma (A2) or 2 years before (A3); A0 = is a title for nurse with five-years higher education degree

Knowledge, attitude and practices’ scores

The KAP means were for knowledge 11.9/19 (63% of max), attitude 11.9/14 (85% of max), practice 10.0/17 (59% of max) (Table 2). The overall Cronbach’s α coefficient was 0.90, with 0.55 for knowledge, 0.90 for attitude, and 0.92 for practice.

Table 2.

Mean score of knowledge, attitude and practice

Variables	n	Mean ± SD
Knowledge score	877	11.9 ± 2.6
Attitude score	877	11.9 ± 2.9
Practice score*	877	10.0 ± 5.4

*Practice is the reported practice

Among participants, most of the respondents (80.6%, 707/877) incorrectly believed that 50–60% of daily caloric intake for patients with diabetes should come from carbohydrates (Additional file 3); many mistakenly identified HDL as the “bad” cholesterol (73.5%, 645/877); and incorrectly believed that the quantity of carbohydrates consumed in a meal is more important than their quality (70%, 612/877). Additionally, many others (67%, 589/877) wrongly stated that patients with diabetes should not exclude any nutrient from their diet. However, most of the healthcare providers correctly stated that animal fats should be limited for diabetic patients (81%, 708/877); correctly stated that diabetic patients should consume fruits (84%, 734/877) (Additional file 3).

Most healthcare providers demonstrated a positive attitude toward dietary recommendations for patients with diabetes. For example, many (89.6%, 809/877) agreed that nutrition and lifestyle are foundational elements of diabetes management (Additional file 4). However, the majority (69.6%, 610/877) agreed that the nutritional management of hospitalized patients with diabetes is solely the responsibility of the nutritionist (Additional file 4).

Healthcare providers’ practices were largely inadequate in several key areas (Additional file 5): coordination with nutritionists when patients modify their diet (58.3%, 511/877); use of a record form to assess nutritional needs (56.9%, 499/877); follow-up with nutritionists for counselling (56.4%, 495/877); and documentation of nutrition-related problems or diagnoses in medical records, such as inadequate dietary diversity, overweight or obesity, undernutrition, or other diet-related barriers requiring nutrition intervention (50.9%, 446/877). Although, counselling on hygiene and dietary measures was widely practiced, with most of healthcare providers (84.5%, 741/877) demonstrating adequate performance in this area (Additional file 5). The most frequently recommended lifestyle changes included increased consumption of fruits, vegetables, and fish, advised by only 55% of providers. Fewer than half (42%) recommended regular physical activity, and just 30% advised patients to reduce alcohol intake and quit smoking (Fig. 2).

Fig. 2.

Lifestyle measures recommended by healthcare providers to their patients

Knowledge score were higher among male providers, doctors, those who have a close relative living with T2DM, those working in public HF (vs. private or faith-based HF) and those who had participated in a recent in job training session (within the past 12 months). Attitude and practice scores were also higher among male providers, doctors, and those who had participated in a recent training. Years of working experience was not significantly correlated with knowledge and practice but negatively correlated with attitude (Table 3).

Table 3.

Mean score for knowledge, attitude and practice according to healthcare providers’ characteristics

Characteristics	n = 877	mean +/- SD	p
Knowledge’s score
Sex
Male	483	12.4 +/- 2.3	0.000
Female	394	11.2 +/- 2.7
Professional categories
Nurse/Nu	610	11.4 +/- 2.7	0.000
Doctor	267	13.1 +/- 1.9
Having close relative with DM
No	550	11.7 +/- 2.6	0.040
Yes	327	12.1 +/- 2.5
Types of health facility
Public	307	12.3 +/- 2.5	0.001*
Private	430	11.6 +/- 2.7
Faith-based	140	11.8 +/- 2.4
Experience (n, β)	877	0.001	0.967
Participation in job training on diabetes the past 12 months
No	645	11.6 +/- 2.7	0.000
Yes	232	12.8 +/- 2.1
Attitude’s score
Sex
Male	483	12.2 +/- 2.7	0.001
Female	394	11.6 +/- 3.2
Professional categories
Nurse/Nu	610	11.6 +/- 3.2	0.000
Doctor	267	12.7 +/- 2.0
Having close relative with DM
No	550	11.9 +/- 2.9	0.526
Yes	327	12.0 +/- 2.9
Types of health facility
Public	307	12.3 +/- 2.2	0.013**
Private	430	11.8 +/- 3.1
Faith-based	140	11.5 +/- 3.7
Experience (n, β)	877	-0.027	0.032
Participation in job training on diabetes the past 12 months
No	645	11.8 +/- 3.1	0.001
Yes	232	12.5 +/- 2.3
Practices’ score
Sex
Male	483	10.5 +/- 5.3	0.012
Female	394	9.5 +/- 5.6
Professional categories
Nurse/Nu	610	9.5 +/- 5.6	0.000
Doctor	267	11.3 +/- 4.6
Having close relative with DM
No	550	10.2 +/- 5.3	0.164
Yes	327	9.7 +/- 5.6
Types of health facility
Public	307	10.6 +/- 5.1	0.079
Private	430	9.7 +/- 5.5
Faith-based	140	9.8 +/- 5.7
Experience (n, β)	877	-0.015	0.532
Participation in job training on diabetes the past 12 months
No	645	9.5 +/- 5.5	0.000
Yes	232	11.6 +/- 4.8

*The Bonferroni test showed public vs. private: mean difference = 0,71 (p = 0,001) and not significant for other groups; **Bonferroni test: mean difference public vs. faith-based = 0.81 (p = 0.020), not significant for the groups

Spearman correlation analysis revealed significantly positive correlations between knowledge and attitude (r = 0.32, P < 0.001) and between attitude and practice (r = 0.49, P < 0.001) (Fig. 3).

Fig. 3.

Correlation between knowledge, attitude and practice

The SEM revealed the pathways linking knowledge, attitudes, and dietary practices (Table 4). Overall, the total effects indicate that both knowledge and attitudes significantly influenced dietary practices. Attitudes exert the strongest direct effect (β = 0.81, 95% CI [0.72, 0.91], p < 0.001), while knowledge also contributes meaningfully (β = 0.54, 95% CI [0.39, 0.69], p < 0.001). The model accounted for 27.2% of the variance in dietary practice scores, suggesting that dietary attitudes and knowledge are critical determinants of dietary counselling practices among healthcare providers. Decomposition into direct and indirect pathways revealed key mediating relationships. Knowledge had a significant direct effect on both attitudes (β = 0.33, 95% CI [0.23, 0.44], p < 0.001) and practices (β = 0.27, 95% CI [0.13, 0.41], p < 0.001). However, its indirect effect through attitudes was equally important (β = 0.27, 95% CI [0.19, 0.35], p < 0.001), confirming that the influence of knowledge on practices is largely mediated by attitudinal change.

Table 4.

Relationship between knowledge, attitude and practices and exogenous factors (sex, professional category, having a close relative with DM and recent training)

Model paths	Total effects		Direct effects		Indirect effects
	β (95% CI)	P	β (95% CI)	p	β (95% CI)	p
Factors affecting Knowledge
Type of HF → Knowledge	-0.16 (-0.41, 0.09)	0.207	-0.16 (-0.41, 0.09)	0.207	-	-
Sex → Knowledge	-0.56 (-0.93, -0.19)	0.003	-0.56 (-0.93, -0.19)	0.003	-	-
Prof. Category → Knowledge	1.33 (0.97, 1.69)	0.000	1.32 (0.97, 1.69)	0.000	-	-
Relative with DM → Knowledge	0.14 (-0.20, 0.48)	0.425	0.14 (-0.20, 0.48)	0.425	-	-
Recent training → Knowledge	0.65 (0.28, 1.01)	0.001	0.65 (0.28, 1.01)	0.000	-	-
Factors affecting Attitude
Knowledge → Attitude	0.33 (0.23, 0.44)	0.000	0.33 (0.23, 0.44)	0.000	-	-
Type of HF → Attitude	0.05 (-0.37, 0.48)	0.804	0.11 (-0.31, 0.52)	0.610	-0.05 (-0.14, 0.03)	0.229
Sex → Attitude	-0.24 (-0.63, 0.14)	0.215	-0.06 (-0.43, 0.31)	0.752	-0.19 (-0.32, -0.05)	0.008
Prof. category → Attitude	0.83 (0.48, 1.18)	0.000	0.39 (0.04, 0.74)	0.029	0.44 (0.25, 0.63)	0.000
Relative with DM → Attitude	0.05 (-0.07, 0.16)	0.427	-	-	0.05 (-0.07, 0.16)	0.427
Recent training → Attitude	0.49 (0.08, 0.90)	0.045	0.28 (-0.12, 0.67)	0.167	0.21 (0.07, 0.35)	0.003
Experience → Attitude	-0.02 (-0.05, 0.00)	0.090	-0.02 (-0.05, 0.00)	0.090	-	-
Factors affecting Practice
Knowledge → Practice*	0.54 (0.39, 0.69)	0.000	0.27 (0.13, 0.41)	0.000	0.27 (0.19, 0.35)	0.000
Attitude → Practice	0.81 (0.72, 0.91)	0.000	0.81 (0.72, 0.91)	0.000	-	-
Type of HF → Practice	0.00 (-0.37, 0.37)	0.999	-	-	0.00 (-0.37, 0.37)	0.999
Sex → Practice	-0.11 (-0.94, 0.73)	0.799	0.24 (-0.48, 0.97)	0.517	-0.35 (-0.71, 0.01)	0.056
Prof. Category → Practice	1.23 (0.49, 1.98)	0.001	0.21 (-0.54, 0.96)	0.589	1.03 (0.66, 1.41)	0.000
Relative with DM → Practice	0.07 (-0.11, 0.25)	0.420	-	-	0.07 (-0.11, 0.25)	0.420
Recent training → Practice	1.79 (0.98, 2.59)	0.000	1.21 (0.48, 1.94)	0.001	0.57 (0.19, 0.96)	0.004
Experience → Practice	-0.02 (-0.04, 0.00)	0.090	-	-	-0.02 (-0.04, 0.00)	0.090

Prof. category = Professional category; *Practice is a reported practice; DM = diabetes mellitus

For female sex, a negative total effect was observed on knowledge (β = −0.56, 95% CI [− 0.93, − 0.19], p = 0.003) and a significant negative indirect effect on attitudes (β = −0.19, 95% CI [− 0.32, − 0.05], p = 0.008). These findings indicate that female providers tend to have lower dietary knowledge, which in turn negatively influences their dietary attitude. After checking for possible confounding by the professional category, it appears that the relationship between sex and knowledge is partially explained by professional category (crude β = -1.01, p < 0.001 vs. adjusted β = -0.56, p = 0.003).

Professional category was strongly associated with higher KAP scores. Compared with nurses, doctors exhibited higher knowledge (β = 1.32, 95% CI [0.97, 1.69], p < 0.001), attitudes (total effect β = 0.83, 95% CI [0.48, 1.18], p < 0.001), and practices (β = 1.23, 95% CI [0.49, 1.98], p = 0.001).

Participation in recent training was consistently associated with improved outcomes. Trained providers reported significantly higher knowledge (β = 0.65, 95% CI [0.28, 1.01], p = 0.001), attitudes (total effect β = 0.49, 95% CI [0.08, 0.90], p = 0.045), and practices (β = 1.79, 95% CI [0.98, 2.59], p < 0.001). Importantly, indirect pathways also showed that training reinforced attitudes (β = 0.21, p = 0.003), which in turn improved practices (Table 4). The structural pathways are summarized in Fig. 4.

Fig. 4.

Structural equation model illustrating the associations between KAP scores

All variables included in the model are observed (measured) variables. Causal relationships are represented by single-headed arrows indicating the direction of influence. Standardised path coefficients are displayed along each directional arrow.

The model showed an excellent fit, with an SRMR of 0.013 and an AIC of 25,864.52, indicating that the hypothesised relationships were well supported by the data (Additional file 6). Because the model was nearly saturated with few degrees of freedom, interpretation focused on path estimates rather than global fit indices. A likelihood-ratio test comparing the indirect model (constraining the direct path to zero) with the full model (direct path freely estimated) confirmed a significantly better fit for the latter (LR χ²[1] = 16.70, p < 0.001), supporting the inclusion of the direct effect as statistically meaningful.

Discussion

This study provides the first comprehensive assessment of KAP regarding nutrition management for patients with T2DM in Kinshasa, DRC. Structural equation modelling offers novel insights into how knowledge and attitudes jointly shape clinical practice. We found significant knowledge and practice gaps, with average scores of 63% and 59% of the maximum possible, respectively, and a mediating effect of attitudes on the relationship between knowledge and practice. Professional category, recent training, and sex were major predictors of KAP levels, highlighting inequities in professional preparedness and opportunities for continued education.

The finding of suboptimal knowledge reveals critical weaknesses in providers’ understanding of nutrition management. Misconceptions regarding carbohydrate distribution and lipid profiles were common, echoing observations from other African and global studies [25, 26]. A lack of knowledge regarding the necessary proportion of energy intake may pose a problem when advising patients on their diets. This issue is particularly significant because it is widely accepted that optimal health outcomes require the quality and quantity of food intake to align with the recommendations [8, 14]. Clearly, if healthcare providers have limited knowledge of nutrition, it will affect the quality of the advice they provide to patients. These gaps underline the need for updated, context-appropriate training modules focused on applied dietary counselling rather than theoretical nutrition, particularly for nurses and general practitioners, who constitute the frontline of diabetes care in the DRC.

Despite a generally positive attitude toward nutrition care, almost 70% of respondents believed that nutrition management of hospitalized patients is the sole responsibility of the nutritionist, despite most facilities having no nutritionist available. Similar findings have been reported in South Africa and Mali, where non-specialist providers refer dietary counselling to dietitians or nurses [31, 54, 55]. This reflects a systemic misunderstanding of role-sharing in diabetes management. Given the shortage of nutritionists in DRC hospitals, this perception undermines continuity of care. Developing clear national guidelines for multidisciplinary nutrition support and integrating nutrition tasks into routine medical and nursing protocols would help bridge this gap and promote task sharing across cadres [41]. Task sharing refers to the collaborative distribution of responsibilities and tasks among different individuals or groups to enhance efficiency and effectiveness in achieving specific objectives. In this context, task sharing refers to the rational redistribution of selected nutrition care activities among doctors, nurses, and nutritionists, with appropriate training, supervision, documentation tools, and referral pathways. This is particularly relevant in settings where nutritionists are scarce and nurses provide much of the routine follow-up for patients with diabetes. Evidence from low- and middle-income countries suggests that task-sharing interventions can improve diabetes care processes and outcomes when they are supported by training, supervision, and clear protocols [41, 56].

A low practice score indicates inadequate application of nutrition care in clinical routines. More than half of the respondents did not use any form or document nutrition-related problems or diagnoses in medical records, consistent with prior studies in sub-Saharan Africa [57, 58]. In this study, nutrition-related diagnoses referred to nutrition problems identified during clinical or nutrition assessment that could be improved through dietary counselling, nutrition intervention, or referral to a nutrition professional [59]. The weak integration of nutrition into clinical workflows and the absence of structured documentation systems limit continuity and evaluation of care. Embedding brief, structured tools such as standardized nutrition assessment sheets and counseling checklists within patient records could improve consistency. Implementation research should assess whether these tools enhance patient outcomes, such as HbA1c control and adherence.

Structural equation modelling revealed that attitudes mediate the relationship between knowledge and practice, confirming the findings from similar KAP models in China and other contexts [34–36]. This highlights that knowledge alone is insufficient to change behaviour unless accompanied by positive attitudes and self-efficacy. Therefore, training programs should not only transfer knowledge but also address beliefs, motivation, and perceived professional roles in nutrition counselling. Integrating attitudinal change modules and role modeling into continuing education could help translate learning into consistent clinical practices.

Differences in KAP scores across professional categories and between men and women suggest systemic inequities in training access and role recognition. Doctors performed better than nurses and nutritionists, and women reported lower knowledge and attitude scores than men. Similar patterns have been observed elsewhere, but not with sex [60, 61]. Even if the relationship between sex and KAP scores is explained by the fact that most nurses are female, the observed negative correlation between KAP scores and female sex remains after adjusting as reported in the model (β = −0.56, 95% CI [− 0.93, − 0.19], p = 0.003). This may reflect the fact that in sub-Saharan Africa, women health workers often have less access to training, mentorship, travel for courses, and promotion opportunities because of gender norms, domestic responsibilities and safety concerns, as shown by a multi-country work on health workforce in Africa [62]. Further investigations are needed. However, given that nurses, the majority of whom are female, are the main healthcare providers in primary care, these disparities have the potential to exacerbate inequalities in service quality in a context where all healthcare is paid out of pocket by the patient and hospital care is generally more expensive than primary care. It is vital to ensure that all professional categories have equitable access to continuing professional development. In addition, the introduction of performance-linked incentives and supportive supervision could be effective in strengthening motivation and skills across all professional categories [63–65].

Strengths and limitations

The notable strengths of this study include the large dataset obtained through multi-stage probabilistic sampling across Kinshasa. The use of validated and adapted instruments with stakeholder input, a clear analytical approach using structural equation modelling with mediation pathways, and strong internal consistency for attitude and practice scales further enhanced the robustness of the study.

This study has some limitations that must be noted. The cross-sectional design prevents causal inference; although SEM helps clarify plausible mediation pathways, longitudinal and experimental studies are necessary to confirm the causal mechanisms. Although SEM allowed us to test theoretically plausible pathways, the cross-sectional design prevents confirmation of temporality. Therefore, terms such as “effect,” “pathway,” and “mediation” should be interpreted statistically rather than causally. Longitudinal or intervention studies are needed to confirm whether improving knowledge and attitudes leads to better nutrition care practices [39, 51]. Self-reported practice measures are likely to be biased by social desirability bias. Additionally, the same questionnaire was applied to diverse professional groups, which may have affected the depth of assessment for specialised roles. Whenever possible, future research should cross-reference self-reported data with direct observations, record audits, and patient feedback.

Another limitation of this study is the moderate internal consistency of the knowledge scale (KR-20 = 0.55). This likely reflects the breadth of topics covered by the knowledge items (applied nutrition, macronutrient goals, lipid interpretation, and clinical counselling), the use of dichotomous scoring, and the variable difficulty and discrimination of the items rather than a failure of the measure’s validity. Items showing low discrimination and being not relevant to nutrition were flagged for revision. It is important to note that the knowledge scale demonstrated the expected validity of the concept: it was correlated with attitudes and practice and contributed significantly to the structural equation model. Additionally, the item-level results were consistent and revealed programmatically relevant deficits. Future research should refine the knowledge items (by removing low-discrimination items, adding items by sub-domain, or adopting polytomous scoring) and reassess reliability in independent samples. Nonetheless, this study provides robust, contextually relevant evidence to guide pragmatic, attitude-sensitive interventions aimed at improving nutritional care for diabetes in urban African settings.

Conclusion

This study has shown a substantial gap in knowledge and practices, which are influenced by both knowledge and attitudes. The observed differences in KAP scores across professional categories and sex highlight the need to provide equitable access to continuing professional development and raise questions of equity in terms of training and fairness of care for patients. Despite the limitations of using a cross-sectional design and self-reported data, the study suggests that improving nutritional care for individuals with diabetes in Kinshasa will require interventions that address both the competence of HCPs and the attitudes that shape clinical behaviour. Policymakers should develop pragmatic interventions, combining skills-based training and addressing the attitudes of providers, contextualised guidelines, performance-related incentive mechanisms, regular supervision, and task-sharing protocols for a rapid improvement in practices in urban settings in the DRC. Further research should assess the effects of these types of interventions on glycemic outcomes in patients.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ADA

American Diabetes Association

AIC

Akaike Information Criterion

DM

Diabetes Mellitus

DRC

Democratic Republic of the Congo

GRH

General Referral Hospital

HA

Health Area

HCP

HealthCare Providers

HF

Health facility

HZ

Health Zone

KAP

Knowledge, Attitude, and Practice

KR-20

Kuder-Richardson formula 20

NGO

Non-Governmental Organization

SANRU

“Santé Rurale” [Rural Health

SEM

Structural Equation Modeling

SRMR

Standardized Root Mean Squared Residual

T2DM

Type 2 Diabetes Mellitus

VIF

Variance Inflation Factors

Author contributions

Steve Botomba, Mapatano, and Muyer carried out the study, participated in data collection, and drafted the manuscript. Pierre Akilimali and Steve Botomba performed the statistical analysis and contributed to study design, data interpretation, and manuscript drafting. Christiane Horwood participated in data acquisition, analysis, and interpretation, and drafted the manuscript. Aurore Beia and Merveille Lukadi, contributed to data interpretation and manuscript drafting. Betty Mizele and Nickson Poka contributed to the conceptualization and manuscript drafting. All authors read and approved the final manuscript.

Funding

No funding was received for this study.

Data availability

The dataset was shared as a de-identified dataset and codebook are available at OSF (https://doi.org/10.17605/OSF.IO/U8P57). Do.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki. The study was approved by the ESPK Ethics Committee (Reference: ESP/CE/190/2024), dated 2 December 2024. All participants received a written information sheet outlining the study’s purpose, procedures, voluntary nature of participation, and data confidentiality measures. Written informed consent was obtained from all respondents prior to data collection. Where participants were unable to provide written consent, verbal consent was documented in the presence of a witness, in accordance with the approved protocol. All collected data were anonymised and stored on password-protected systems accessible only to the study team. The study aimed to improve diabetes care services by identifying gaps in provider knowledge, attitudes, and practices, and guiding future interventions.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.