Design
This will be a two-armed, open-masked, parallel group, stratified cluster randomized controlled trial (RCT) of a NUCOD program (Figure 1), with health center as a unit of randomization. We will use a computer-generated list of random numbers to randomize health centers stratified by Kathmandu University (KU) outreach centers and Government health posts, whereby all patients from the same health center will be allocated to the same group. The study will be planned and implemented in concordance with the Consolidated Standards of Reporting Trials (CONSORT) cluster trial extension statement [8] and Standards for reporting implementation studies (StaRI) statement [9]. We plan to use the cluster design to reduce between-group contamination and spill-over effects of the intervention as well as to align the design with the natural project implementation unit based on the cluster.
< Figure 1: Summary of trial design for NUCOD program>
Setting
The study will be conducted in the 26 clusters- 9 outreach centers (Bahunepati, Manekharka, Hindi Health, Baluwa, Bolde Phediche, Dapcha, Kartike Deurali, Salambu, and Dhunkharka Health Centers) of Dhulikhel Hospital, Kathmandu University Hospital and 17 Government health posts located in the adjoining catchment areas will be purposively selected from the Kavrepalanchowk and Sindhupalchowk districts. The primary stratification variable, as well as the unit (cluster) of randomization and implementation, will be health clinics (outreach centers and health posts [HP]). There will be 13 clusters per stratum (1:1 allocation ratio). Outreach centers are the small medical health facilities that offer basic and emergency care services in remote areas of Nepal. These centers are run by the Kathmandu University Hospital. Similarly, health posts are the Government run facilities that provide basic health care and preventive medication in remote rural areas of Nepal.
Kavrepalanchowk district located in Province 3 of Nepal consists of 13 municipalities; 6 urban and 7 rural municipalities with a total of 137 wards. The headquarter of this district is Dhulikhel municipality. A total of 381,937 population resides in this district; 62.51% live in urban and 37.49% live in rural municipalities. There is one 15-bed District Hospital, 4 Primary Health Care (PHC), and 86 HPs in the district [10]. Sindhupalchowk district also located in Province 3 of Nepal consists of 12 municipalities; 3 urban and 9 rural municipalities with a total of 103 wards. The headquarter of Sindhupalchowk district is Chautara. A total of 287,798 population resides in this district; 41.42% live in urban municipality and 58.58% live in rural municipality. There is one 15 bedded District Hospital, 3 PHC, and 75 HPs in the district [10]. The map of Kavrepalanchowk and Sindhupalchowk district with the distribution of outreach centers and health posts are shown in Figure 2 and Figure 3.
<Figure 2: Map of Kavre District with Outreach Centers of Dhulikhel Hospital and Government Facilities. ORC= Outreach Center>
<Figure 3: Map of Sindhupalchowk District with Outreach Centers of Dhulikhel Hospital and Government Facilities. ORC= Outreach Center>
Participants
Eligibility criteria for health centers/health posts
Eligibility criteria for participants
To resemble the real-world settings as much as possible, we will apply the minimum inclusion and exclusion criteria for the participants. Anyone (i) with confirmed pre-existing type II diabetes and prediabetes at the time of our screening (i.e., pre-existing diabetes or prediabetes); or (ii) with confirmed diagnosis of type II diabetes or prediabetes after our community screening process (i.e., newly diagnosed diabetes and prediabetes); and who (iii) is not planning to relocate outside of the current place of living in next 2 years; and (iv) is older than 18 years of age, will be eligible for project recruitment. People will be excluded if they (i) are not psychologically capable of communication; and (ii) are diagnosed as type 1 diabetes. Participants will be assessed for the risk of diabetes using the Indian Diabetes Risk Score (IDRS) [11]. Anyone identified as at-risk will undergo random finger-prick glucose test followed by the hemoglobin A1c (HbA1c) test. The diagnosis will be confirmed by HbA1c level of ≥ 6.5% for diabetes or between 5.7%-6.4% for prediabetes [12].
Screening and Recruitment
The program will be rolled out with each outreach center/health post serving as the implementation unit. The program participants will be recruited through a community-based screening effort. Information regarding the screening event will be disseminated at the time of awareness campaign. Trained diabetes nurses will screen the general population at the outreach center/health post, and those screened positive will be referred to the DH diabetes clinic for further tests and for program recruitment if eligible for the study. All the screening procedures including lab investigations and anthropometric measurements will be carried out at the respective centers, and all the longitudinal data of patients will be entered in an Electronic Health Record (EHR) system for data storage and management. Each participant will be assigned a unique identification number at the time of enrolment. All data collected as part of this study will be identified with this number. Research assistants (RAs) will be responsible for obtaining the signed informed consent from the participants.
Recruitment will be primarily through conducting diabetes awareness mass campaigns in the targeted sites of the two districts to achieve adequate participant enrolment. We intend to reach 50,000 people (approximately 2000 people at each of the 9 outreach centers and 17 health posts). Interested candidates will contact study nurses who will provide them basic information about participation, including information on the screening process, time commitment and expectations associated with participation. Assuming 25% of those who are made aware of diabetes will participate in the screening (see below for the details of screening methods), we target to screen 12,500 participants. The two-step screening approach will be used for the study after assessing the risk of diabetes among the individuals by using the IDRS assessment tool. In the first step, anyone under risk will undergo random finger-prick blood glucose test using Beurer glucometers [13]. A random blood sugar level of 200 mg/dL or higher will be considered as positive for diabetes and sugar level of 140mg/dL to 199 mg/dL will be considered as positive for prediabetes [12]. In the second step, participants screened positive for a random blood glucose level will undergo a glycosylated hemoglobin (HbA1c) test. The HbA1c level between 5.7% to 6.4% will be considered as positive for prediabetes and HbA1c level of 6.5% or above will be considered as positive for diabetes [12]. Participants with HbA1c level ≥ 5.7% will be invited to participate in the study. Based on earlier quoted prevalence of 8.4% (95% CI: 6.2% - 10.5%) for diabetes [1] and 13% (95% CI: 11.8% - 14.5%) for prediabetes [14], we expect to identify 1050 (12500*0.084) individuals with diabetes and 1625 (12500*0.13) individuals with prediabetes for this study.
Randomization
The 26 health clinics will be randomized 1:1 into the continuum of care group (intervention) and the usual care group (control), stratified by outreach center and health post. Usual care group are the ones who continue managing their diabetes under the direction of their primary care provider, which is the most commonly used method for diabetes management in Nepal. Selection of usual care group as a comparator is thus justified. A statistician otherwise not involved in the project will perform this group assignment at cluster level with simple random selection with the statistical program R. Randomization will be done before participant recruitment. Outcome assessors will be blinded to the group assignment, but the nurses and patient participants will be aware of their group assignment. If blinding to the outcome assessor is accidentally broken, they will follow a standard protocol for re-assessment of the patients at another time. Any violations of the study protocol will be recorded and reported to the Ethics Committee.
Baseline Assessment
At baseline, trained RAs will interview the participants using a standardized electronic questionnaire using CommCare, mobile platform designed for data collection [15]. RAs will receive two weeks of training on data collection and ethical issues. The questionnaire will assess socioeconomic characteristics including age, sex, ethnicity, religion, marital status, annual income, education level, family history, and lifestyle factors including smoking, alcohol intake, and physical activity. We will use the Global Physical Activity Questionnaire [16], and calculate the metabolic equivalent of task (MET) minutes per week. A weekly MET equivalent of 600 would represent 30 minutes of brisk walking five times per week or 15 minutes of running five times per week. We will use Prime Screen questionnaire [17], a short diet assessment tool to assess the diet quality of the study participants. Body weight will be measured with minimum clothing and without shoes using an Omron Model HBF-400 scale and recorded to the nearest 0.1 pounds. The weighing scale will be calibrated to zero every day. Participants’ heights will be measured, without shoes, while the participants stand against a wall. Height will be measured using a tape measure and recorded to the nearest 0.1cm.
Blood samples will be collected for HbA1c, low-density lipoprotein (LDL), high-density lipoprotein (HDL), triglycerides, and total cholesterol at each respective health center where all the laboratory procedures will be carried out. Blood samples will be collected using evacuated blood collection tubes. Participants will be asked to fast overnight (8-14 hours).
Intervention: NUCOD Program Components
The intervention will be delivered by specially trained nurses from the Dhulikhel Hospital (DH). The intervention has been developed based on our extensive literature review and the recommendations from the Disease Control Priorities-3 as well as an analysis of local contexts in consultation with various stakeholders. The intervention will include the following components, logically sequences as a continuum of care approach. Figure 4 presents the major program components. The nurses will act as the leader and coordinator for the implementation of all those program components.
-
Training of nurses: By drawing from standard training manuals published from International Diabetes Federation and American Diabetes Association, the project management team will first develop a protocol and content for the training of nurses for this program. This training protocol will be tailored to the context of Nepal and will include case studies and examples relevant to Nepal. The training will focus on clinical as well as case management skills. We will train 26 nurses (one for each site) into diabetes-nurses using this training protocol.
-
Community awareness campaigns: The nurses will lead the organization of a mass-campaign on diabetes in coordination with local newspapers, radio stations, youth-groups, municipality offices, district health offices, and other health facilities. The campaigns will, in particular, include a simple diabetes risk factor assessment tool to be adapted from the Type 2 Diabetes Risk Test (DRT) of the American Diabetes Association [18]. The campaign will encourage people with high risk to attend the screening program.
-
Screening programs: The nurses will organize screening programs for both diabetes and pre-diabetes for the community members on specific dates and sites which will be communicated beforehand to the people in the locality. The screening will be done in coordination with the clinical biochemistry department of DH.
-
Linkage to clinical Care: People screened positive for diabetes and prediabetes (pre-existing as well as newly) will be linked to the DH clinic by nurses where they will be recruited into the program and will receive their group assignment. People assigned to the control group will follow the usual procedure, while those assigned to the intervention group will attend a one-hour nurse-led counseling session on diabetic care. The patients will be responsible for their medical costs (lab investigations and medicine), while people experiencing financial hardship can apply for charity care (free or at a reduced price) with guidance from the nurse. The assigned nurse will be responsible for determining their financial profile by reviewing the annual household income, applicable assets, available insurance coverage, and confirmation of other sources of payment.
-
Community follow-up counselling and support to the diabetic patients: The nurse will organize diabetic patients in groups of 10 and arrange bi-monthly meetings (each two-hour long) in which the patients will be facilitated using the 5 A framework (Assess, Advise, Agree, Assist, and Arrange follow-up) to adopt healthy lifestyle (dietary changes, physical activity, cessation of smoking, etc.) [19]. In addition to the lifestyle changes, the participants will also be facilitated to adhere to medical advice (timely follow-up, adherence to medications, etc.). The nurses will also coordinate with the diabetes clinic at the hospital to ensure that the clinical care and the community-level interventions complement each other. They will also have an electronic tablet that has all the details (laboratory results, behavioral parameters, anthropometric measures, etc.) of the participants so that the progress of the participants in these parameters can be recorded and assessed.
-
Prevention programs for pre-diabetic participants: In addition to the people with diabetes, the nurses will also form groups of prediabetes participants and will arrange monthly meetings to help facilitate the adoption of healthy lifestyle using the Diabetes Prevention Program (DPP) curriculum. The DPP lifestyle intervention has been associated with significantly reducing the development of diabetes through its structured behavior changing approaches [20]. A multidisciplinary team of dieticians and physical therapists will assist with the other components of the DPP program that are related to nutrition and physical activity.
The control group will receive the diabetic usual care, which includes the same hospital quality improvement for diabetic care as mentioned above, the follow-up services from the Female Community Health Volunteers (FCHVs) (but without the nurse coordination and supervision as in the intervention group) in addition to the usual care. FCHVs are the grass root level community health workers in Nepal who play important roles in the implementation of community-based programs including, screening for high-risk cases, referring people to local health facilities, and maintaining a record of health activities [21]. The program comparison of the intervention versus the control is summarized in Table 2.
<Figure 4: Intervention components of the NUCOD program. DH= Dhulikhel Hospital>
Outcome measures
We will examine both clinical outcomes and implementation outcomes as we hypothesize that the nurse-led continuum of care program will improve both health and the implementation of priority interventions in diabetes. The outcome measure will pertain to the individual participant level. The primary outcome for the diabetes group is glycemic control (normal glucose level in blood) as measured by HbA1c level (4% - 5.6%) at 12 months from the intervention. The HbA1c level is indicative of the average level of blood glucose over the past 2 to 3 months [22]. The primary outcome for the prediabetes group is the incidence of diabetes at 12 months from the intervention.
The secondary outcomes are blood pressure, body max index (BMI), and lipid levels. Also, the secondary outcomes include a series of implementation outcomes in the RE-AIM implementation framework (Reach, effectiveness, adoption, implementation, and maintenance) [23-25]. We provided details on those measures below (see Table 1 for a summary).
Clinical outcomes. The clinicians from Dhulikhel hospital outreach centers not otherwise associated with the project will measure HbA1c, lipid profiles, BMI and blood pressure when the participants visit the diabetic clinic at baseline, 6 and 12 months.
The HbA1c will be measured using Boronate affinity chromatography (Axis-Shield, Norway) [26]; LDL and HDL using the elimination method (Dialab, Austria) [27]; triglyceride using GPO-PAP (Dialab, Austria) [28]; and total cholesterol using CHOD-PAP (Dialab, Austria) [29]. For each type of assay, the laboratory has quality control (QC) materials (using commercially available assayed and unassayed control material) from Bio-Rad Laboratories, USA. Each QC is run at least in duplicate. External QC is arranged by internationally recognized reference laboratories that distribute batches of samples of various concentrations for each assay. The laboratory performs the External Quality Assurance Scheme from an unknown assayed sample from the Department of Clinical Biochemistry CMC, Vellore, India for 23 routine parameters, 5 immunological parameters, and HbA1c. Additionally, 5% of the blood samples will be obtained in duplicates and sent for testing all parameters, blinded to the laboratory personnel.
The mean of three measurements of systolic and diastolic blood pressure, using a Microlife automatic blood pressure measuring device, will be adopted for analysis (mmHg). Hypertension is defined as systolic blood pressure ≥140 mm Hg or, diastolic blood pressure ≥90 mm Hg; or taking antihypertensive medication [30]. Weight will be measured and recorded to the nearest 0.1 kg, without shoes, and with minimum clothing, using an Omron Model HBF-400 scale. Height will be measured without shoes using a standard tape measure with participants standing against a wall for measurement and recorded to the nearest 0.1cm. Body Mass Index (BMI) will be calculated as weight in kilograms divided by height in meters squared. Overweight is defined as a BMI of 25 kg/m2 or higher and obesity is defined as BMI of 30kg/m2 or higher based on international cut points [31].
Implementation outcomes. RE-AIM: (1) Reach will be measured by the number of people participating in the program divided by the number of people eligible to be recruited into the program; (2) Effectiveness will be represented by the clinical outcomes; (3) Adoption at the patient level will be measured by the proportion of people adherent to the clinical advice in lifestyle and self-care. These will be measured by the self-reported Summary of Diabetes Self-Care Activities (SDSCA) scale [32] at baseline, 6 months and 12 months. The SDSCA measure is a brief self-report questionnaire that includes items assessing general diet, specific diet, exercise, blood-glucose testing, foot care, and smoking. In addition, the Diabetes Treatment Satisfaction Questionnaire (DTSQ) [33] will be administered at baseline, 6 and 12 months as well. DTSQ is the most commonly used patient-reported outcomes in diabetes trials, which reflects the patient perception of the treatment. Adoption at the clinic level will be measured by a number of participants served using health records and examining nurses’ adoption of intervention program through interviews; (4) Implementation will examine fidelity and quality of the program execution per protocol. To check program fidelity, we will select and train standardized patient (SP) from the program participants with stable conditions. Those SPs will serve as the “secret agents” and the sentry to assess program fidelity with a quality checklist through their routine encounters to the clinicians at the DH diabetes clinic and the nurses. The clinicians will be blinded to the status of the SPs. The development of SP and checklist will follow the protocol we have developed in a separate study [34]. (5) Maintenance will not be assessed for the purpose of this study.
The hospital and program administrative system will collect a range of other information including detailed program costs, health service utilization, and incidence of complications and comorbidities. All data will be entered and securely stored in EHR, a secured online data capturing, and management system developed for the study at the DH.
Sample size
In this proposed community-based intervention, the changes in the absolute reduction of HbA1c levels (continuous outcome) is being considered for calculating the sample size. We aim to detect a clinically significant reduction in HbA1c levels from 7.6% to 6.5% (SD=1.5%) (based on our assumption) among the individuals with diabetes during the 12 months period [35, 36]. Setting the statistical significance at the 0.05 level, seeking 90% power, assuming an intraclass correlation coefficient (ICC) of 0.01, an average cluster size of 50 (including individuals with diabetes and prediabetes) based on the preliminary data from awareness and screening campaign, design effect for clustering of 1.49, and design effect for unequal cluster size of 1.52 (coefficient of variation of cluster size 0.25), 91 people with diabetes per arm (182 total) will be required for the study. For the incidence of diabetes among people with prediabetes (dichotomous outcome), we considered the incidence of 10% and 2% respectively for the two groups; stated as a range in the earlier study [37]. Based on the above considerations, 471 people with prediabetes per arm (942 total) will be required. We expect a drop-out rate of 10% from baseline to 12 months follow up; hence, we will need to recruit a total of 200 diabetes and 1036 prediabetes participants (total sample size= 1236 participants). The sample size for our cluster RCT was calculated using the equation below. Detail of the sample size estimation is available in another study conducted by Ribeiro et al. 2018 [38].
SScluster RCT = SSstandard RCT X DEcluster X DEunequal
where,
SScluster RCT = total sample size for cluster RCT;
SSstandard RCT = total sample size for standard RCT (equal to (Z1−α/2+Z1−β)2 2σ2 / Δ2) where Zx is the x’th percentage point of the standard normal distribution, Δ is the clinically important difference in treatment means and σ2 is the variance in the outcome)
DEcluster = design effect for clustering (equal to 1 + ICC*(m-1) where m is the number per site)
DEunequal = design effect for unequal cluster size (equal to 1 + [(1 + cv2) X m − 1]*ICC) where cv is the coefficient of variation of cluster size and m is the mean cluster size)
Statistical analysis
Data will be analyzed at the individual (patient) level. Statistical analyses will be performed using the intention-to-treat (ITT) approach in the originally assigned groups. Demographic and baseline characteristics for the intervention and control group will be presented in the form of mean (standard deviation [SD]) or 95% confidence intervals (CIs) for continuous variables and counts (percentages) for categorical variables. ITT analysis will be performed on the final data collected at 12 months.
Generalized estimating equations (GEE) with clustering by site, an exchangeable correlation matrix, and robust variances will be used to assess program effect adjusting for the potential baseline covariates. The model will take account of clustering and will be used to look at the difference between groups for HbA1c level at 6 and 12 months respectively. Multiple imputation will be used to account for the missing values assuming they are missing completely at random. To test for the secondary sensitivity analysis, data will be analyzed without multiple imputation and without baseline covariate adjustment. The difference in the incidence of diabetes between the groups among the individuals with prediabetes will be tested using a GEE. Statistical significance will be assessed at the 5% level and all analysis will be 2-sided. All data analyses will be performed using Stata 15 (StataCorp, Texas, USA) statistical software program.
As we expect the program effect may differ between several subgroups of the program participants, we plan to do secondary analyses of the following three subgroups for the primary (HbA1C) and secondary outcomes (incidence of diabetes): (i) the subgroups with poor initial glycemic level (HbA1c ≥ 5.7%) (people with poorer glycemic level at baseline may have a stronger desire to change their behavior); (ii) the subgroups divided by gender (in Nepal, women tend to be more adherent to clinician guidance); (iii) the subgroups of different social economic classes (people in higher socioeconomic class may be more likely to adhere to lifestyle changes); and (iv) government vs. non-government health centers. The secondary analyses will compare the HbA1c level and incidence of diabetes among people with prediabetes among these subgroups at the end of the 12-month intervention using the generalized linear mixed models with a random cluster effect and adjusting for potential confounders.
Economic Analysis: Cost-effectiveness analysis will also be conducted with a Markov simulation model to be developed. The Markov model has been widely used to describe the development of the chronic non-communicable diseases [39]. All direct costs (e.g., screening cost, cost of training of nurses, and cost of EHR system development) and indirect costs will be collected. All costs will be reported in 2020 US dollars using the exchange rate of 1 USD = 111.7 NPR. We will solve the model numerically over a short and long timeframe (i.e., 10 years and 20 years), and calculate the incidence of diabetes, costs and quality-adjusted-life-years (QALYs) under the control case or NUCOD. Incremental QALY and costs of NUCOD relative to the control case will be calculated with an annual discount rate of 3%. We will use WHO-CHOICE (Choosing Interventions that are Cost-Effective) thresholds for cost-effectiveness: an intervention is defined as "cost-effective" if it produces a healthy life year for less than three times gross domestic product (GDP) per capita, and as "very cost-effective" if it produces a healthy life year for less than the GDP per capita [40, 41].