Cost of diabetes mellitus and associated factors – an institutional cross-sectional study in Ghana

Abstract

Background

Diabetes mellitus, like many other chronic diseases, is costly to manage and poses a substantial economic burden on individuals directly and indirectly. In this paper, we studied the associations between cost of diabetes and socio-demographic characteristics.

Methods

This was a cross-sectional cost-of-illness study that employed systematic random sampling. We collected data from 385 respondents at the Tamale Teaching Hospital of Ghana between June and August of 2023. Prevalence-based costing and the human capital approach were employed to arrive at total cost of illness. Regression analysis was used to find associations between sociodemographic characteristics of the respondents and the total cost of illness.

Results

The mean total cost of diabetes mellitus per year is $290.44. Mean direct annual cost of illness per year is $159.70 representing 54.99% of the total cost while the mean indirect annual cost per patient is $130.72. Being male (B = 0.42, 95% CI 0.02–0.82; p = 0.039), living in an urban area (B = - 1.05 95% CI - 1.58 – - 0.53; p = 0.000), having a longer duration of illness (B = 0.04, 95% CI 0.003–0.07; p = 0.032), and having the complications of diabetic retinopathy (B = 0.42, 95% CI 0.02–0.82; p = 0.041) and stroke (B = 1.26, 95% CI 0.52–2.00; p = 0.001) were statistically significant in association with total cost of illness.

Conclusions

Various demographics with diabetes carry different dynamics in terms of cost burden. We recommend a tailored approach to care for individuals with diabetes mellitus and their families as a protection against catastrophic health care expenditure that could result from a high cost of illness.

Introduction

Diabetes mellitus (DM), also known simply as diabetes, is a chronic disease of the endocrine system that develops when there is either a relative or absolute lack of insulin secretion by the pancreas [1]. 10.5% of adults worldwide have DM, and it is estimated that 6.7 million adults died in 2021 from the disease or its complications [2]. 24 million adults are estimated to have diabetes in Africa; in 2021, about 416,000 people in Africa died from diabetes and this is projected to keep increasing [3]. In Ghana, prevalence of diabetes is between 2.0% and 6.5% [4,5,6,7].

Despite the increasing prevalence of diabetes and the complications that typically follow, the healthcare spending in lower- and middle-income countries (LMICs), aside being inadequate, is still generally more concerned with communicable diseases [4], thus leaving the absorption of cost of care of diseases like diabetes to individuals and their families; this financial impact can be quite substantial, and it comes in direct and indirect terms [8, 9].

Direct costs comprise cost of medications and investigations (direct medical cost) and transportation and food costs during hospital visits (direct non-medical cost), while indirect costs are the wages an individual loses as a result of illness from diabetes or diabetes-related hospital visits.

Globally, cost of DM ranges from $29.91 to $740.10 per patient per year [10,11,12,13]. In Ghana, although literature remains scanty, the mean monetary cost of diabetes care is reported to range between $372.65 and $1,207.80 [14, 15].

Individuals with diabetes are more likely to experience catastrophic medical costs than those without diabetes [16], nevertheless, very little is known about the factors that influence the cost of illness. It has been shown that cost of DM differs by sociodemographic characteristics: studies conducted in countries in Africa, Asia, and Europe have found factors such as sex, type of area of residence (rural versus urban), and presence of complications to be associated with a higher cost of DM [11, 13, 14].

In this study, we first estimated the total cost of illness and then determined the associated factors that influenced this cost. This study provides further insight into the limited knowledge about the cost of illness among people living with DM in Ghana against a backdrop of underdevelopment, poverty and a relatively weak healthcare system, and what particular demographics have a higher cost of illness.

Methods

Study design and setting

This study employed an institutional cross-sectional design and collected data from the Tamale Teaching Hospital (TTH) between June and August of 2023. TTH is the third largest hospital in Ghana, serving as a specialist referral hospital for all of northern Ghana and beyond, seeing patients from neighbouring Togo and Burkina Faso. TTH runs a specialist out-patient department (OPD) diabetes clinic that sees an average of 2,532 new diabetic cases per annum.

Ethical considerations

Formal permission was sought from, and granted by, the TTH before the study commenced. Ethics clearance was obtained from the Committee on Human Research, Publications, and Ethics of the Kwame Nkrumah University of Science and Technology, Ghana (No. CHRPE/AP/473/23) before data collection. Informed consent to participate was obtained from all of the participants. Respondents were assured of the utmost level of confidentiality and privacy before taking part in the study. Participants also retained the right to withdraw at any time without prior notice, and with no consequences whatsoever.

Sampling and data collection

Patients with DM were recruited using a systematic random sampling technique. There is a register for patients with DM that have been booked for visits at the clinic. A sample size of 387 was calculated using the Taro Yamane formula.

$$\:n=\frac N{1+Ne^2}$$

(1)

$\:n$ = sample size

N = the population size (N = 2910; the yearly average number of new diabetic visits to TTH OPD over the last two years since the electronic health management system was fully adopted)

e = acceptable sampling error (e = 0.05)

This gave a sample size of 351.7. A non-response rate of 10% was assumed, and so a final sample size of 386.9 (rounded up to 387) was determined for this study.

The list of patients booked to be seen between June and August 2023, containing 581 individuals, was used as the sampling frame. We calculated the sampling interval, k, by dividing 581 by the sample size, 387. k was calculated as 1.5 and rounded up to 2 which then became the interval of selecting subsequent study participants after randomly selecting the first participant for the day. Diabetes clinic runs three days in a week, and so the selection process continued on each consecutive diabetes clinic day until a sample size of 385 was obtained, representing a response rate of 99%. A structured questionnaire (Supplementary file 1) uploaded to Google electronic forms was used to collect data ranging from participants’ sociodemographic characteristics, health history, to cost of illness (i.e., direct and indirect). The data were collected in languages the respondents opted to use (English, Twi, or Dagbani) by trained research assistants who were proficient in all three languages and then uploaded into the electronic forms.

Study variables

Socio-demographic data such as age, religion, area of residence (which was later on dichotomized into rural/urban based on classifications by the Ghana Statistical Service [17]), and occupation were self-reported by the respondents, while data on complications and co-morbidities were extracted from the electronic health management system (EHS). Total cost of illness was calculated as the sum total (in Ghana Cedis) of the direct cost and indirect cost. Respondents were required to estimate the amount of money they spend on medications monthly and this was multiplied by 12 months to obtain that particular expenditure per year. They were also required to state the number of DM clinic visits they had in the last year, and their estimates for consultation costs, investigations, transportation, and food were multiplied by this number of visits to get the total amount per year. Furthermore, hourly earnings based on gross household income was obtained from participants, and these data were used to calculate how much a respondent lost based on hours off work.

Inclusion and exclusion criteria

Diabetes patients with a diagnosis of at least a year who were attending the TTH either for regular follow up or as a referral case from other health facilities were included in this study. This is in line with the prevalence-based cost-of-illness approach which typically involves measuring the total costs of illness or disease within a specified period such as a year [18].

Pregnant women with diabetes and non-pregnant individuals with diabetes who required urgent or emergency treatment such as those presenting with hypoglycaemia or diabetic ketoacidosis were excluded.

Data analysis

Data were analyzed with Microsoft Excel 2021 and Statistical Package for the Social Sciences (SPSS) v27 after being checked for completeness. Total cost of illness was estimated in two major categories: direct cost and indirect cost. Direct costs were estimated over the past year through respondent recall. It was computed by summing up the cost of drugs, investigations, transportation to the health facility, and food and water during hospital visits. Indirect costs were determined by estimating the hourly wage of the respondent, and then placing a monetary value on the hours they spent away from work due to illness and hospital visits with an assumption that the time of all employed respondents had a monetary value to them, and that anyone who was not working had none. In line with the human capital approach, we further assumed that a person who was visiting the hospital for diabetes-related care at any point in time or away from work due to diabetes illness was losing earnings or wages.

Monetary data were collected in the Ghanaian Cedi and then converted to the United States Dollar using the August 2023 Bank of Ghana exchange rate of 1 Ghanaian Cedi to 11.01 United States Dollars (USD). We then generated dummies for all the independent covariates and regressed them against total cost of illness as the main dependent variable and transformed the results for robust standard errors. The study results report beta coefficients as magnitude with statistical significance or otherwise (alpha values set at p < 0.05).

Results

Socio-demographic and clinical characteristics

Table 1 shows the socio-demographic and clinical characteristics of the study respondents. This study had a total of 385 respondents (n = 385), representing a response rate of 99%. Female respondents were in the majority, 266 (69.1%). Respondents were predominantly in the 55–64 age bracket expressed as 116 (30.1%). In all, 176 (45.7%) had no formal education, while 94 (24.4%) had tertiary education. Of the total number of respondents, 249 (64.7%) were employed, with 103 (26.8%) stating that they engaged in trading and 84 (21.8%) being civil servants. A large percentage of respondents (86.6%) resided within an urban area. Majority of the study participants (65.7%) were Muslims. Of the respondents, 152 (39.5%) had at least one complication or comorbidity. Overall, 121 (31.4%) had a duration of illness of less than five years, 103 (26.8%) had a duration of illness of 5–10 years, with the rest, 161 (41.8%) having an illness duration of over ten years. The range duration of illness was 33 years with a lower quartile of 4 years, median of 7.5 years, and an upper quartile of 11 years.

Table 1 Socio-demographic and clinical characteristics of study participants (n= 385)

Total cost of illness

The mean total cost of illness of DM per patient per year is GH¢ 3,197.57 ($290.44) (Table 2). The mean direct cost of illness per patient per year is GH¢ 1,758.34 ($159.70) representing 54.99% of the total cost of illness while the mean indirect cost per patient per year is GH¢ 1,439.23 ($130.72), representing 45.01% of the total cost.

Table 2 Total cost of illness of diabetes mellitus

Total cost of illness and associated independent covariates

Table 3 shows the association between socio-demographic characteristics and total cost of illness. Being male is associated with a higher cost of illness as compared to being female (B = 0.42, 95% CI 0.02–0.82; p = 0.039), participants who lived with a diagnosis of diabetes for 10 or more years have a higher cost of illness (B = 0.04, 95% CI 0.003–0.07; p = 0.032), living in an urban area is associated with a lower cost of illness as compared to those in rural areas (B = − 1.05 95% CI − 1.58 – − 0.53; p = 0.000), and having the complications of diabetic retinopathy (B = 0.43, 95% CI 0.02–0.83; p = 0.040) and stroke (B = 1.26, 95% CI 0.51–2.01; p = 0.001), is associated with a higher cost of illness as compared to those with no complications. There was no statistically significant association between factors such as age, education, marital status, employment status, and comorbidities on total cost of illness. The table below shows the summary of the findings.

Table 3 Association between total cost of illness and respondents’ sociodemographic characteristics

Discussion

The total cost of DM was found to be substantial against the backdrop of a part of sub-Saharan Africa that is further impoverished and underdeveloped [11, 12, 19], and considering the fact that the minimum wage in Ghana in 2023 was $1.35, and that many of the people who live in northern Ghana are not high-income earners, financial difficulties may arise from the management of this disease, leading to a detrimental effect on the ability of an individual to be adherent to medication and treatment guidelines.

This study found that being male was associated with having a higher cost of illness, and this is in agreement with a study conducted in India [20]. However, a number of previous studies conducted in Vietnam, Bangladesh, and the Netherlands found that being female was associated with a higher cost of DM [11, 21, 22]. While a study conducted in India found no association between sex and cost of DM [13], the likely explanation for the cost being higher in males as found by this study is that, in Ghana, men are more likely to be employed, educated [22], and to earn more than women, and so they are likely to be able to afford medications and also lose more income when they are away from work as a result of the illness, leading to a higher overall cost.

We found an association between place of residence and cost of diabetes: an individual with DM living in an urban area has a lower cost of illness than one living in a rural area. This finding is inconsistent with studies conducted in Ethiopia and India [23, 24]. Generally, urban residents are economically stronger as compared to their counterparts in the rural areas and are more likely to access healthcare in the urban areas which is within reachable distance, and although current trends in prevalence studies show that prevalence is directly correlated with urbanization [25, 26], individuals living in urban areas possess better access to preventive healthcare and do not have to travel long distances to access diabetes-related care.

The presence of complications drives the cost of illness up [11, 22, 27, 28], and this study found that those particularly with complications of stroke and diabetic retinopathy were more likely to have a higher cost. Specifically, stroke, one of the cardiovascular complications of DM, results in a higher cost of treating DM [29]. In addition, diabetic retinopathy has been found to drive up the direct medical costs of DM in a study that was conducted in Singapore [30]. In general, complications of any chronic disease, particularly diabetes that can have several long-term sequelae, result in the need for more medications, hospital visits, a higher likelihood of frequent admissions, and more days off work due to illness or physician appointments. All of these factors lead to a greater likelihood of having a higher cost burden.

Studies conducted in Ethiopia, India, and Ghana [11, 14, 27] have found associations between having a longer diagnosis duration and incurring a higher cost of illness. These studies have shown that the healthcare-associated costs of diabetes increase over time. This study has found that association as well. It stands to reason that an individual who has had diabetes for more than ten years is more likely to have complications and be on multiple pharmacological agents for better blood glucose control, driving cost up.

The current analysis did not find any associations between cost of illness and other socio-economic and clinical factors. Although studies have found that a highly educated individual was more likely to be faced with a higher cost [22, 24], other studies have found that having higher education was associated with a lower cost of DM [13, 21], while in a household where the head had lower education, there was likelihood of catastrophic healthcare spending in relation to DM [31]. Furthermore, it can be postulated that having other diagnoses alongside the disease, such as hypertension, might predispose to a higher economic burden, and although a study conducted in Ethiopia [23] found associations between comorbidities and duration of illness with total cost of illness, this study found none. Finally, a systematic review [32] as well as studies in Malaysia [33] and Ghana [34], concluded that being religious was associated with improved glycaemic control in those with diabetes, however, even if that is the case, that might not necessarily translate to a lower cost of diabetes [35], as this study has found, there is no association between religion and cost of illness.

This study has some strengths and limitations. The study’s strengths lie in the fact that it estimated both direct and indirect costs, providing a more holistic appreciation of cost that is not confined to just what is spent on medication and hospital visits. This study also has some limitations which need to be addressed by future research. First of all, lack of records to differentiate types 1 and 2 DM may result in findings applied generally to both major types of DM when that may not be the case. Furthermore, the authors did not study adherence to treatment among the respondents. An individual who is not adherent to treatment in the first place would not buy prescribed drugs as often and would consequently incur a lower direct cost and indirect cost (since such an individual might have less hospital visits). This might lead to an underestimation of cost of diabetes. Finally, since respondents were required to report what they estimated to be their expenditure on certain items as well as time spent in hospital and off work, there was a risk of recall bias which could affect the results of the study.

Conclusions

This study found that being male, having a longer duration of diagnosis and having complications of DM (specifically stroke and diabetic retinopathy) were associated with a higher cost of illness, while living in an urban area was associated with having a lower cost of illness, despite the fact that adherence, which could influence cost of illness, was not studied. Different demographics carry different dynamics, and any interventions to alleviate the cost-of-illness burden should be sensitive to these differences. A tailored approach that is research-driven is required to develop policies that ensure that individuals and their families are protected from catastrophic health care expenditure that can result from a high cost of illness. Doing so will provide a more nuanced approach in ensuring that the cost-of-illness burden is reduced in an equitable manner. Finally, a conscious improvement of the healthcare system of Ghana, where prevention and access to primary care are prioritized, is necessary for first of all averting DM, or at least providing timely and adequate primary care that results in less likelihood of patients developing complications that may drive cost up.