Boosting diabetes and pre-diabetes detection in rural Ghana

List of abbreviations

IDF International Diabetes Federation

WHO World Health Organization

CHPS Community-based health planning and services

BMI body mass index

CDP confirmed diabetic participants

USP unknown diabetic status participant

NCD non communicable disease

Introduction

Diabetes is one of the fastest growing non-communicable killer diseases in the world, claiming one life every eight seconds and a limb every 30 seconds¹. Diabetes of all types can lead to complication in many parts of the body and can increase the overall risk of dying prematurely^2–7. Pre-diabetes condition occurs when blood sugar levels are higher than normal, but are not high enough to be classified as diabetes; this often has no symptoms, and is reversible³.

According to the latest 2016 data from the World Health Organization (WHO), amongst adults living with diabetes melilites, 80% live in low and middle-income countries such as those in the Asia and Eastern Pacific region. The largest number has been reported in China (90 million people⁶), followed by India (61.3 million people) and Bangladesh (8.4 million people)⁵. Complications of diabetes results in increased morbidity, disability, and mortality and have a high economic cost, especially in developing countries⁸. More specifically, the reported prevalence of type 2 diabetes ranges from 1% in rural Uganda to 12% in urban Kenya. While gestational diabetes has been reported in Sub-Saharan countries at varying levels (e.g. from 0% in Tanzania to 9% in Ethiopia⁹). Lastly, even considering those values an underestimate, it is expected that the reported cases will reach 82 million by the 2030¹⁰.

Ghana is challenged with the increasing prevalence of diabetes, similar to other African countries, with a prevalence of 3.6% in adults and 518,000 diagnosed cases within the country¹¹. More specifically, The prevalence of diabetes in some parts of Ghana has been found to be higher than the world average of 6.4%^12,13. Moreover, the 2015 report of the IDF indicated many other cases probably remain undiagnosed, posing an increased danger of complications for people living with diabetes unaware of the consequences. Previous studies in the country showed that low level of physical activity and obesity were associated with increased risk of diabetes⁴. Additionally, old age and level of education were also associated with increased risk of diabetes⁴. It has also been observed that within Ghana sugary drinks consumption is linked to type 2 diabetes¹⁴. Community-based health planning and services (CHPS) is a national health program in Ghana adopted in 1999 to reduce geographical barriers to health care access¹⁵. According to the CHPS policy, relocating nurses directly to communities could outperform an entire sub-district health center. The cost-effectiveness of CHPS for malaria, diarrhea, and pneumonia has been recently reported¹⁶. However, specific interventions for non-communicable diseases such as diabetes have not yet been investigated within the CHPS policy.

Vulnerable populations such as those in low- and middle-income countries are generally more affected by diabetes related complications¹⁷. As in several fields of healthcare, mobile health (mHealth) has the potential of reducing for vulnerable populations with diabetes¹⁸. This can occur either by sending reminders or by increasing access to patient management¹⁹. Despite the plethora of studies on mHealth and diabetes management^19–21, no study has been carried out in rural Africa with the aim of improving detection of diabetes and pre-diabetes by using mobile technologies and community nurses.

This study proposes a novel approach based on community nurses using glucometers and mobile phones, performing tests on undiagnosed and diagnosed subjects proactively within the community without waiting for participants to present at the clinic. The main objective is to develop a novel method to increase diabetes and pre-diabetes detection, additionally we aim at finding new behavioral determinants related to those conditions. In particular, the purpose of the mobile app is to simplify the tracking operations of the nurses and to collate the data into a centralized secure server. For this purpose, a pilot project was carried out in rural communities of the Central Region of Ghana to assess the feasibility of the approach. A secondary objective was to look for new behavioral determinants related to the rural Ghanaian populations by carrying out a comparison with a group of subjects diagnosed with diabetes from the same communities. Similar to a project carried out in the same area about improving prenatal care²², the project utilized community nurses instead of the participants to assure reliable glucose level data collection.

Methods

Data source

The research was conducted in the central region of Ghana, specifically within the Biriwa and Anomabo communities. These two communities are in the Mfantsiman Municipal District and based on the 2010 census, the two towns have a population of about 7,500 and 14,389 respectively. Those communities have been chosen due to them being relatively close to a main road connecting urban centers. The assumption is that the members of those communities are more prone to adopt unhealthy habits (such as smoking and drinking) which are more common in urban centers. These communities are also used to interacting with community nurses and there is a general level of trust. However, people still show restraint from several types of tests due to the additional costs not covered by basic health insurance.

The community nurses collecting the data were equipped with glucometers and low-cost Android smart phones. They received a short (less than one day) training on the app and were supported on its use during the first week of the project. Data were stored through the mobile phone app and sent to a server to improve management and facilitate eventual longitudinal screening. The developed app was based on the CommcareHQ framework, and comprised a series of guided questions that the nurses completed in addition to the glucose test (questions used are available as Extended data²³). Those guided questions are an extension of existing standardized questionnaires as the FINDRISK test²⁴. Some screen-shots of those questions are depicted in Figure 1. If network was not available at the point of data collection, information could be sent later when the network was available. CommcareHQ is a popular mobile data collection, and it has been used in several projects. For a review on projects using the CommcareHQ platform, the reader is addressed to 25.

Figure 1. Screen-shots showing some of the guided questions the community nurse were completing while performing the glucose tests.

The questions related to diet were based on general consumption for a typical week and not the week at point of sampling.

Figure 2 shows the typical two steps of the screening, first a nurse is performing a glucose test (on the left), and then the data are recorded through the app (on the right). Collected data from respondents were from the rural communities, both male and female, including pregnant women with the intent of capturing eventual gestational diabetes case²⁶. The sampling of candidates at risk was based on physical factors and at the nurse’s discretion. During regular visits to the rural communities the nurses assessed whether the person to be tested was a potential diabetic candidate and therefore deemed to be tested (e.g. appearing overweight), paying particular attention to pregnant women. By doing this we have introduced a piloted bias (as overweight people, pregnant women and other cases were deemed at risk by the nurses). Therefore, with those biases, the resulting statistics is not representative of a random screening of the national population.

Figure 2. A typical two steps screening, first a nurse is performing a glucose test (on the left), and then the data are recorded through the app (on the right).

Nevertheless, our focus is not to estimate total incidence of undetected diabetes or pre-diabetes, but to propose an approach that can detect as many as possible cases which otherwise would go unnoticed.

However, to avoid a strong bias individual deemed healthy that were willing to be tested were included. Any detected diabetic or pre-diabetic cases were immediately informed, and lifestyle changes or pharmaceutical therapies discussed. Data were collected on known diabetic patients as well, to evaluate the differences or similarities in lifestyle among the various groups.

With the mobile phone app, the community nurses could also keep track of longitudinal changes or whether a subject has been already tested, in a similar manner of a project conducted in the same area about boosting prenatal care²². Subjects who came into contact with community nurses were asked the last time they had a meal and based on the response blood glucose measurements were classed as either fasting blood sugar levels or random blood sugar levels. Furthermore, the following information were recorded from the subject a) Anthropometric indices (weight, height, BMI) measured by the nurses b) Demographic Information (sex, age) self-reported c) Blood glucose measurement (fasting or random sugar test) measured by the nurses d) Information on risk factors (pregnancy, family history of diabetes, history of hypertension, kidney disorder, alcoholism, low levels of exercise and unhealthy eating habits) self-reported. OneTouch Ultra (Milpitas, CA, USA) and Accu-Check (Hoffmann–La Roche SA, Basel, Switzerland) strips and glucometers were used to measure the blood glucose level. According to the WHO classification²⁷, a fasting blood glucose level greater 110 mg/dL but less than 126 mg/dL was considered pre-diabetic. A fasting blood sugar level 126 mg/dL or above was considered diabetic. A random blood sugar level greater 140 mg/dL but less 200 mg/dL was considered pre-diabetic. Above 200mg/dL was considered diabetic. Family history of diabetes is referred to occurrence of diabetes in close relatives defined as either father, mother, siblings or offspring. Most of the data were assessed by the nurse though some were self-reported by the subjects and this might represent a limitation. The data entered through the developed Android app were stored in a secure server provided by Dimagi. Data were entered in the forms of the mobile app and transmitted, encrypted, to the cloud-based server, where they were accessed and downloaded via a password-protected web interface.

Results

The diabetes screening saw over 204 inhabitants in Anomabo and Biriwa over a period of 6 months. Of those, 103 were previously confirmed diabetic participant (CDP) with an average age of 62.9 ±11.2 years, and 101 people, with an average age of 30 ±9.7 years, with unknown diabetic status participant (USP). Originally 211 forms were completed, however, 7 of them were deemed erroneous during the quality check and therefore were removed before the analysis. For each person data were collected only once. The CDP cohort comprised 66 female and 37 male subjects, while the USP cohort comprised 95 female and 6 male subjects (see Underlying data²³). Details of the main demographic characteristics revealed for the two groups are reported in Table 1 while the breakdown of the study variables is distributed in the other tables. Community nurses see on average 20 patients per day 5 times a week. Their duties mainly encompass malaria, diarrhea, and pneumonia treatments which are generally perceived as more urgent. During the pilot, the total number of people approached for the diabetes screening was 240 with a participation rate of 88%. The 12% of who refused the screening reported as the main motivation the unwillingness to sign the written consent for the study.

Two-sample t-test performed comparing the BMIs was statistically significant (P value <0.05) although they were both normoweight. Some subjects of the CDP cohort also presented with co-occurrence of ulcers (n=4), asthma (n=1), arthritis or rheumatism (n=3), and kidney disease (n=1). At time of testing participants in the USP cohort presented with co-occurrence of asthma (n=4), arthritis and rheumatism (n=1), and typhoid fever (n=1). The subject presenting typhoid fever showed a random glucose blood level of 132 mg/dL which could not be considered neither diabetic nor pre-diabetic. Therefore, typhoid fever was not considered a confounding factor as the tested participant did not show a high value due to this. These results are summarized in Table 2.

Discussions

Ghana is challenged with the increasing prevalence of diabetes, which is similar to that of other African countries. The increase in diabetes in Ghana—along with other non-communicable diseases (NCD) like stroke—is part of an epidemiologic transition. Increase of socioeconomic status seems linked to increase of NCDs incidence. However, recent surveys have shown that employment and education play a role in this context, suggesting that rural communities which are in proximity to urban centers are prone to increased unhealthy habits leading to NCDs, and are more vulnerable than urban centers²⁸.

Mobile phone apps and pro-active screening can help community nurses to spot new cases of diabetes and pre-diabetes. In particular, the proposed approach was based on pro-actively performing blood screening during rural visits of the community nurses, who were collecting information via the mobile phone app. This can also help tracking and monitoring, as the nurses can follow up the status of the participants during future visits.

Our findings indicate that some individuals in vulnerable populations, such as those in rural Ghana, are not aware of becoming diabetic or being in diabetic condition. We report two cases of diabetic participants in the USP cohort (2%) and 20 as pre-diabetic (19.8%), which we consider to be high when compared to previously reported statistics²⁹. However, it must be taken into account that the cohort selection was not purely random, and piloted bias was introduced as the subjects to be tested were chosen according to the nurse’s discretion. Therefore, these percentages should not be taken as a representative sample of the national population. The proposed approach aims instead at detecting as many cases as possible of diabetics and pre-diabetics which otherwise would go unnoticed. For this purpose it proved to be successful, inexpensive and easily integrated into the standard duties of community nurses.

Initially the nurses were equipped with ihealth glucometer dongles for the smartphone (ihealth, Mountain View, CA, USA). The anecdotal comments from nurses were that. despite the initial interest, they were not practical to collect data, though they might be suitable for a single individual. The reason can be related to the familiarity of the nurses to known tools such as Accu-Check and OneTouch, or the cumbersome use of switching continuously between the mobile app of the glucometer dongle and the app to record the data. The use of a mobile app can allow for large institutions to easily monitor diabetes in rural areas through the collected data held in a secure centralized server. Moreover, mobile apps appear more user friendly for the nurse in comparison to cumbersome paperwork.

During the qualitative interviews, one nurse pointed out that despite the marginal costs of glucose tests both patients and government are not promoting them, while it could be cost-effective for Ghanaian institutions to detect pre-diabetic cases instead of dealing with a growing diabetic population, as has been shown for similar vulnerable populations³⁰. Emphasis on pre-diabetes has been recently criticized in US and Europe due to the related possible speculations³¹. As indeed, there is currently the possibility for pharmaceutical companies of seeing an opportunity and funding research on pre-diabetes and promoting unnecessary demands. Nevertheless, healthy eating and regular physical activity have been shown to delay or prevent progression to diabetes³² and this will be particularly helpful in low and middle income countries. Indeed, in the specific case of Ghana, it will not be beneficial to promote medications addressing pre-diabetes. What was gathered from the qualitative investigation is the need of further investigations related to the necessary physical activity to maintain an healthy lifestyle and prevent diabetes, which would be inexpensive, but requires social campaigns.

Despite the progresses made in Ghana to achieve universal health access and coverage, financial barriers to diabetes service utilization still exist. Subjects usually do not undergo glucose tests due to financial constraints, as the test is not covered under most standard medical insurance policies, as is the case in Nigeria and Tanzania³³. One nurse mentioned that subjects were often not able to report their dietary habits and were unaware of the effects of their diet on their health. Conversely to prenatal care³⁴, it appears that generally the members of the communities trust the personnel of the clinics for this type of tests.

At population level the CDP and USP cohort were normoweight (having a BMI between 18.5 and 24,99), though there were subjects which were obese and with hypertension in both groups. It is worthwhile to mention that the CDP subjects might have some dietary changes already after being informed of being diabetic, which could have affected their weight and other measurements. However, from the interviews it seems that the major changes were the reduction of consumption of salt and sugar, some increased their consumption of vegetables and fruit, and some reducing the consumption of alcohol beverages. The study did not track behavioral changes, but we can presume that these may have occurred and can represent confounding factor. No statistically significant difference in alcohol consumption between the two groups was detected. However, this could be due to the fact that some individuals in the CDP cohort changed their dietary habits (10 people reported to have cut out alcohol consumption after their diagnosis). Moreover, the nurses were instructed to focus for both groups on women which might consume fewer alcoholic beverages than men. Nevertheless, the general impression of the nurses was that the alcohol consumption had an impact. With the increase in quality of life in the country, western habits such as alcohol consumption might be also increasing, and therefore augmenting the risk of diabetes. It has been noticed that people informed of their condition tend to change their dietary habits. Nevertheless, given the country-wide growing trend in alcohol consumption, social marketing campaign related to this³⁵ should be considered.

Almost all subjects of both groups reported to perform regular physical activities related to their job. However, this information seems vague and it is not clear whether this physical activity is aerobic or resistance or whether it is sufficient to keep a healthy glucose level. Most likely further activities should be proposed. Jogging and other sport activities are inexpensive and easy to promote. Therefore, promoting this type of sport activities can address this issue. Moreover, it appears necessary in the future to use more detailed investigations about sport activities, such as using the WHO global physical activity questionnaire³⁶.

Ghana has experienced an exponential increase of the mobile network, social media and smartphones in the recent years³⁷. Beyond the screening of the population carried out by nurses, smartphones can have an impact on glycemic control, as smartphone dongles can be inexpensive and attractive to young users. Strategies such as gamification, and social media should be explored to increase awareness on glycemic levels as shown in other contexts³⁸.

Among the limitations of the study, there is the small sample size and the heterogeneity of the used samples. These were mainly dictated by the funding of the study and it is suggested that the study is repeated with larger sample sizes. Moreover, we focused on early detection and dietary habits, and their relation to cardiovascular diseases has been only investigated through the assessment of past medial history of hypertension.

Conclusions

Proactive glycemic screening on vulnerable population – such as those living in rural areas – can be effective in detecting new cases of diabetes and pre-diabetes. Our approach using community nurses screening subjects deemed at risk and collecting data on mobile phone was found to be effective, and suitable for longitudinal studies. Campaigns increasing awareness of alcohol consumption, physical activity, nutrition and healthy habits should be emphasized in any prevention strategy as the population seems to still be unaware of the consequences.

Despite this, studies with larger population are required to confirm the results. The diabetes and pre-diabetes screening described in this manuscript can be easily included into the national CHPS policy with several potential benefits. Those benefits include reducing incidence by detecting cases of pre-diabetes which hopefully will not convert into type-2 diabetes and enabling timely treatment of diabetes patients avoiding complications related to delays in treatment.

Ethics and consent

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Written consent for the reported data was collected. The Noguchi Memorial Institute for Medical Research of the University of Ghana recorded the study with the identifier 076/13-14.

Data availability

Software availability

The CommCare software is required to use the source code, questions used in the app are available as Extended data

Software available from: https://play.google.com/store/apps/details?id=org.commcare.dalvik&hl=en

Source code available from: https://github.com/alecrimi/diabetesUP/tree/v1.0

Archived source code at time of publication: http://doi.org/10.5281/zenodo.2587117²³

License: Creative Commons Zero “No rights reserved” data waiver