Diabetes management in Thailand: a literature review of the burden, costs, and outcomes

Management of diabetes represents an enormous challenge for health systems at every level of development. The latter are tested for their ability to continuously deliver high quality care to patients from the day they are diagnosed throughout their life. In this study, we review the status of diabetes management in Thailand and try to identify the key challenges the country needs to address to reduce the current (and future) medical and economic burden caused by the disease. We conducted a literature review on the burden, costs, and outcomes of diabetes in Thailand. This information was complemented by personal communication with senior officials in the Thai Ministry of Health. We identified the following priorities for the future management of diabetes in Thailand. First, increasing screening of diabetes in high risk population and promoting annual screening of diabetes complications in all diabetic patients. Second, identifying and addressing factors affecting poor treatment outcomes. Third, policy should specify clear targets and provide and use a monitoring framework to track progress. Fourth, efforts are needed to further improve data availability. Up-to-date data on the medical and economic burden of diabetes representative at the national level and at least the regional level are essential to identify needs and monitor progress towards established targets. Fifth, promotion of a healthy lifestyle for prevention of diabetes through education and quality information delivered to the public.


Background
Thailand is an upper-middle income country in South-East Asia which has achieved impressive economic, social, and health improvements in the past ten years. Between 2007 and 2011 its gross domestic product grew by 7.8%, the percentage of people living below poverty line decreased from 21% to 8.1% between 2000 and 2009, and under-five mortality reached a record low for a country with 13 deaths per 1000 live births in 2010 [1].
The country's disease burden and demographic profile stand out from the South East Asia region's average for its high prevalence of non-communicable diseases (NCDs), higher than the average regional prevalence of HIV/AIDS, and an ageing population [2].
Non-communicable diseases are estimated to account for 71% of all deaths with cardiovascular diseases representing 27%, cancers 12%, and diabetes 6% of all deaths [3]. Communicable diseases are estimated to account for 24% of the total years of life lost (regional average 49%) while noncommunicable diseases account for 55% (regional average 36%) [3].
Life expectancy in 2009 was 70 years (66 for males and 74 for females) [4]. The total fertility rate has decreased from 2.4 in 1990 to 1.6 in 2006 [5]. In terms of age structure (based on 2009 figures), 22% of the Thai population was below age 15 and 11% above age 60, the population median age was 33. Many factors have contributed to these remarkable health achievements and among these, increased access to health care through universal health insurance is probably one of the key ones. Thailand was the first country in the region able to achieve universal health coverage of its population. Universal health coverage was achieved in 2002 after 27 years of progressive expansion of the universal coverage (UC) scheme to include all the uninsured together with the beneficiaries of the former medical welfare scheme for the poor, the elderly, the disabled and the children, and the voluntary health insurance scheme for the informal sector [6]. This scheme is financed through general taxation and covers 75% of the Thai population [7].
The two other main public schemes are the civil servants medical benefit scheme, financed through general tax and a non-contributory scheme, and the social security scheme which is financed by equal contributions from employer, employee and the government [8]. In 2010, these two schemes covered 9% and 16% of the population respectively [8].
In terms of the benefit package, the UC scheme covers for in-patient and out-patient care, prescription drugs, laboratory investigations, simple dental care procedures, disease prevention, health promotion, several expensive treatments such as radio-and chemotherapy, surgical procedures, and emergency treatment [7]. Since the end of 2008, renal replacement therapy including renal and peritoneal dialysis and kidney transplantation is also covered by the UC scheme although a co-payment per haemodialysis session is required [9].
Very little information is available on medicines availability in Thailand. According to a survey using the standard WHO/HAI methodology, in 2006 metformin was available in more than 80% of the surveyed public (20 hospitals) and private (21 private retail pharmacies) health facilities [10]. Diabetes medicine such as metformin, sulfonyurea and insulin are available to all diabetes patients free of charge as part of the UC scheme.
Management of diabetes represents an enormous challenge for health systems at every level of development. The latter are tested for their ability to continuously deliver high quality care to patients from the day they are diagnosed throughout their life. Achieving this requires good coordination across different areas of health care, different levels of care, in addition to trained human resources, an efficient supply system for medicines, a reliable health information system, national and international policies and strategies, and an equitable financing system which ensures access to essential health care services.
In this study, we review the status of diabetes management in Thailand and try to identify the key challenges the country needs to address to reduce the current (and future) medical and economic burden caused by the disease. To do that, we review evidence on the burden of diabetes, including the available data sources, available diagnostic and screening programmes, diabetes treatment, costs and outcomes, and policies implemented.

Methodology
This study is based on secondary data analysis complemented by primary data collection. A literature review of peer-reviewed and grey literature including policy documents and annual reports of the MoPH and government statistics was undertaken. The following key words were used on the 21 st  ). Studies presenting data on prevalence, incidence, mortality, outcomes (testing, screening rates), and complications were included. We excluded studies which did not present disaggregated data on diabetes (e.g. studies on chronic disease including diabetes). Evidence from the systematic literature review was complemented by primary data obtained through personal communication with Thai diabetes experts and senior public health officials from the Bureau of Non-Communicable Diseases (NCDs) and the Bureau of Policy and Strategy in the Thai Ministry of Public Health (MoPH).

Results
The search strategy yielded 267 papers, of which 194 were excluded through title and 37 through abstract screening because they did not meet the inclusion criteria. There were 46 final peer-reviewed papers included in the review (Table 1).

National health examination survey (NHES)
There is no national longitudinal diabetes registry in Thailand and the main source of prevalence data for diabetes is the national health examination survey (NHES). The aim of the NHES is to estimate the prevalence of particular health conditions and risk factors including obesity, diabetes, and mental, reproductive and elderly health. This survey was completed for the fourth time in 2009 (previous surveys were conducted in 1991, 1997, 2004). In 2009, a national representative sample of 20,450 (39,290 in 2009) individuals aged 15 and over was randomly selected using a four-stage sampling strategy from five provinces in each of the four regions and Bangkok. Diabetes prevalence was assessed by fasting blood glucose test and patients identified as diabetic if they had either FPG > = 7.0 mmol/L but lack of a previous diagnosis (undiagnosed diabetes) or a previous diagnosis of diabetes and intake of glucose lowering drugs during the past two weeks (diagnosed diabetes) [11,12].  [16].

Other studies
Several other studies have been conducted on the burden of diabetes in Thailand. These include investigations on the incidence [17,18] and prevalence of diabetes in population sub-groups [19], prevalence [20,21] and incidence [22][23][24][25][26] of diabetes Type 1 in children, incidence [27] and prevalence [28][29][30] of gestational diabetes, and prevalence of type 2 diabetes in women with polycystic ovary syndrome [31]. Other studies looked at the prevalence of complications in the Thai diabetes population.

Prevalence and incidence
Prevalence of diabetes mellitus type 2 Data from the four NHES indicate that prevalence of diabetes in individuals aged 15 and over has increased over time from 2.3% in 1991 to 4.6% in 1997, to 6.8% in 2004, and to 6.9% in 2009 [32] (Figure 1). According to NHES 2009 and 2004, women experienced a higher prevalence than men [11,12]. In contrast, findings from the InterASIA study in 2000 did not identify any difference in diabetes prevalence between men and women aged 35 and over (9.3% men vs. 9.9% women, p = 0.6) [16]. At the other end, an earlier study a among Shinawatra employees, a group of relatively young individuals with high socioeconomic status, found a higher prevalence in men (2.2%, N = 1,250) than in women (0.1%, N = 2,365) [19]. According to the NHES 2009, prevalence of diabetes was higher in urban areas in comparison to rural areas (P < 0.001 for both sexes) [11]. However, findings from NHES 2004, only found a statistically significant higher prevalence in urban men, in comparison to their rural counterparts, (P < 0.001) but not in urban women (p > 0.05) [12]. The existence of a difference in diabetes prevalence between urban and rural areas in Thailand was confirmed by findings from the InterASIA study in 2000 (12.1% urban vs. 8.4% rural, p = 0.01, in individuals aged 35 and over) [16]. No difference between urban and rural areas was identified in the InterASIA study in 2000 and the NHES survey in 2004 [12,16].
Different studies on diabetes prevalence in Thailand agree that the prevalence of diabetes increases with age and reaches a peak at some point after age 55 depending on the study [11,12,16,19].
An important problem in diabetes care is under-diagnosis as it delays start of treatment and exposes the patient to the risk of complications which leads to higher treatment costs. Levels of under-diagnosis improved between 2004 and 2009 and this improvement was more evident in women than men. Nevertheless, a large proportion of total diabetes patients remains undiagnosed (from 66.5% to 47.3% in men and from 51.4% to 23.4% in women between 2004 and 2009) [11].
Findings from the NHES 2009 suggest that women have better rates of diagnosis in comparison to men and this difference was statistically significant (there was a difference also in 2004 but this was not statistically significant) [11,12]. However, the InterASIA study did not find a statistically significant difference in the proportion of diagnosed patient by gender (53% women vs. 47% men, p = 0.4) but it found that diagnosis rate was higher for people aged 55 and over (63% > =55 years old vs. 37% <54 years old, p = 0.01) [16].
A comparison of the findings from the NHES in 2009 with results from 2004, shows that the proportion of individuals with diabetes and concomitant hypertension did not significantly decrease in 2009 in both sexes [11]. However, the proportion of women with diabetes who were either abdominally obese or had high total cholesterol (≥5.2 mmol/L) increased from 18% in 2004 to 23.5% in 2009 and this difference was significant (both P < 0.01) [11].
An earlier study on the prevalence of DMT2 in children and adolescents (mean age 11.6 years) reported an increase from 5.8% to 13.3% between 1986 and 1995 [21]. The authors suggest a link between this increase and the concomitant increase in obesity from 5.8% to 13.3% between 1990 and 1996 [21].

Incidence of diabetes mellitus type 1 in children
Incidence of type 1 diabetes in children in North-East Thailand has increased over the years from in 0.17 per 100,000 in 1984 to 0.3 in 1995, to 0.39 in 2000 and 1.27 in 2005. In the other regions it has also grown although following a less linear trend particularly in the Central region ( Figure 2).
Findings from a study in the Bangkok-Noi district in Bangkok, found an extremely small number of cases (maximum 1 per year) between 1991 and 1995 which, despite the size of the child population in this district, it still resulted in the highest prevalence rates for type 1 diabetes recorded in the country (2.18 per 100,000 in 1991, 0 in 1992, 1.97 in 1993, 2.06 in 1994, and 2.04 in 1995) [24].
Another study looked at the seasonal variation of diabetes type 1 and found that the peak seasons were the winter and the summer season and lower during the rainy season [20]. This was explained by the higher prevalence of infections in the winter season and the higher levels of pollution in the summer in comparison to the rainy season [20].
All studies (apart from the Bangkok study which did not discuss gender differences) on diabetes type 1 identified Results from a study in Bangkok were not included as they were not comparable with the other studies due to the different methodology employed [24]. There was another study looking at the seasonal variation in DMT1 which was not included because it did not report incidence rates but only total number of cases in one hospital which made it unsuitable to calculate population incidence rates [20]. Incidence was calculated by dividing the total number of cases reported by hospitals with the total child population in the hospitals' catchment area. Data were obtained from retrospective postal survey or medical record analysis and no information on the diagnostic criteria used was provided. A study summarising the results of the regional studies mentions that the criteria in the 1997 Report of the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus were used. higher incidence in girls, with a girls to boys ratio ranging from 1.3 to 2. However, none of them tested for statistical significance. Peak age at onset was 10-14 in North and North-East region [22,25,26], 11-15 in the South region [23], and 9-12 in Bangkok [24].

Incidence of diabetes mellitus type 2 in adults
We identified three studies on incidence of DMT2 in adults in urban Thailand from a high socio-economic background [17,18,33]. All the three studies used the diagnostic criteria from the American Diabetes Association (ADA) using fasting plasma glucose tests (FPGs), in addition to that one study also used oral glucose tolerance tests (OGTTs). The most recent study was among professionals and office workers in Bangkok and found an incidence rate in the age-group 35-60, of 17.8 per 1000 person-years (PY) in men and 9.2 per 1000 PY in women in 2005 [17]. A second study among university hospital employees in Bangkok reported an incidence of 13.6 per 1000 PY in men and 6.4 per 1000 PY in women between 2001 and 2005. The study participants were over 35 years old and predominantly female (three-quarters). High BMI (>25 kg/m2), elevated FPG (> = 96 mg/dl) and alanine aminotransferase levels (> 18 mg/dl) were shown to be independent predictors of DMT2 [18]. Risk of diabetes in men was approximately twice as high as in women but this apparent association was confounded by higher BMI and FPG levels in men and the crude rates showed there was no association between gender and DMT2 [18]. Findings from this study were supported by an earlier study which found an overall incidence of 11.3 per 1000 PY among employees of a state enterprise in Bangkok between 1985 and 1997 [33].

Gestational diabetes
Four studies on the prevalence of gestational diabetes mellitus (GDM) were identified [27][28][29]34]. They were hospital based (one hospital per study) and mostly used the diagnostic criteria of the National Diabetes Data Group (NDDG), (50 g + 100 g OGTT) with one study also comparing NDDG results with WHO criteria (75 g OGTT). All hospitals were located in Bangkok ( Figure 3).
The highest prevalence (15.7%) was estimated in 1995 for women in their 24 th to 28 th week of gestation using WHO criteria [34]. When using NDDG criteria, the same study estimated a remarkably lower prevalence of 1.4% [34]. Another study highlighted the importance of the timing of diagnosis, 5.3% of screened pregnant women were diagnosed with gestational diabetes before their 20 th gestational week and an additional 4.9% of previously undiagnosed women were identified in a second round of testing during their 28 th to 32 nd week of pregnancy [27]. The 1987-1989 study followed a sub-sample of women after delivery by performing an OGTT using WHO criteria 4-6 weeks after childbirth. Results showed that 42.2% of the 71 women tested had abnormal carbohydrate tolerance, 7% had diabetes and 35.2% impaired glucose tolerance (IFG) [28].
The most recent study (March 2003-January 2005) aimed to assess the percentage of pregnancies with GDM which were missed at the time the study was conducted because clinical guidelines in the study hospital limit screening to women at high risk of developing GDM. Eligibility criteria for screening included pregnant women with at least one of the following risk-factors: aged 35 and over, family history of diabetes, previous birth over 4 kg, and other known risk-factors of GDM. This study showed that the prevalence of women aged 30 to 34 was 5.7%, and that of the 32 cases identified, 12 had one risk factor, one had two risk factors (1). More importantly 18 cases of GDM showed no risk factors and would have been missed due to the recommendations in the hospital's guidelines at the time of the study [29].

Mortality
Results of a comprehensive study to verify causes of death using medical records and verbal autopsy methods, suggest that the leading cause of death in both men and women in 2005 was stroke (9.4% and 11.3% respectively) [35]. Diabetes was the second cause of death in women (8.0%) and the tenth cause of death in men (3.2%) in 2005 [35]. This difference is partly explained by the high mortality levels due to road accidents and HIV/AIDS in men. This study also highlighted issues with vital registration where a high percentage of deaths were classified as illdefined. Adjustments to vital registration data led to important changes in the proportion of deaths due to HIV/ AIDS, ischaemic heart disease, and diabetes [35]. The proportion of deaths attributable to diabetes reaches its maximum at age 50-79. In 2005, they represented the main cause of death in females (12.3%) and seventh in males (5.4%) in this age group [36].

Costs
Using a micro-costing approach, a study including 475 patients receiving treatment in a hospital in North-East Thailand estimated a median cost of illness per patient of USD 140 (mean USD 881) in 2008 (1 USD = 32 THB, at 2008 prices) [37]. The total annual cost included 23% direct medical costs, 40% direct non-medical costs, and 37% indirect costs [37]. Direct medical costs were driven by inpatient service costs (11% of total cost of illness), while direct non-medical cost were dominated by the cost of informal care (28% of total cost of illness). Cost of permanent disability (19% of total cost of illness) was the most important contributor to informal cost [37]. The median cost of illness increased with age (p-value < 0.001), duration of the disease (p-value < 0.001), the level of fasting blood glucose level (p-value = 0.002) and number of co-morbidities (p-value = 0.0013) [37]. The median cost of DMT1 was higher than DMT2 (USD 748 vs. USD 140), however this difference was not statistically significant (p = 0.167 [37]).
The monthly median cost of informal care based on a sample of 190 informal care givers was estimated at USD$ 27 using an opportunity cost approach and USD$ 23 using the proxy good method [38]. Cost of informal care contributed 28% of total cost of illness of diabetes [37] (Table 2).
Findings from one hospital showed that the cost of informal care accounted for 28% of the total cost of diabetes [38]. This is likely to have negative repercussions on labour force participation of informal carers who dedicated on average 112 hours per month to informal care and were in their most productive age (25-54 years old) [38].

Diabetes screening and prevention
Recently, Thailand launched the new "Thailand Healthy Lifestyle Strategy 2011-2020 Plan" endorsed by the Ministry of Public Health [39]. This plan aims to decrease the prevalence, complications, disability, mortality and cost of illness of five major non-communicable diseases including diabetes, hypertension, ischemic heart disease, stroke and cancer. The strategies proposed include: healthy public policy, social mobilization and public communication, community building, surveillance and care system including screening diabetes in high risk population, and capacity building [39].
Yet, to-date, there is still no national screening and no prevention programme in place. There are some examples of successful sub-national diabetes screening and prevention initiatives. These include a prevention model for diabetes in primary care, an educational programme on diabetes prevention for community health care workers, and use of a mobile health unit to screen people living in rural areas for chronic and other health conditions [40][41][42][43][44]. Another example is the development of a simple diabetes risk score to identify Thai patients at high risk of developing diabetes. This score is based on a set of variables (age, BMI, waist circumference, hypertension, and history of diabetes in parents or siblings) which were identified as being significant predictors of diabetes [33]. Measurement of these variables does not require laboratory tests, making the model a cost-effective instrument to identify high risk people to be screened.
However, despite these initiatives have shown initial positive results, they not really been taken beyond the initial pilot phase.

Diabetes treatment
Thailand has its own guidelines for diabetes treatment (only available in Thai) [45] endorsed by the Thai Diabetes Association, the Endocrine Society of Thailand, and the Ministry of Public Health. The guidelines are updated every three to four years and the latest version was published in 2011. More than 5,000 copies have been published and distributed to primary and secondary care physicians in all the regions of Thailand and they are also available online. However, there is no evidence on how many physicians have adopted these guidelines.
The Thai guidelines resemble the WHO/IDF guidelines [46,47], particularly with reference to screening, prevention, treatment and monitoring. However, use of HbA1c is not universally recommended for the diagnosis of diabetes because of the lack of standardization of the HbA1c measurement methodology in Thailand and FPG is recommended instead. HbA1c is still generally recommended for monitoring treatment outcomes.
To control diabetes and reduce the risk of developing complications, it is essential to optimise physiological values such as glycaemic levels and blood pressure and to perform regular examinations for nephropathy (albumin excretion, serum creatine), retinopathy (eye), neuropathy (distal symmetric polyneuropathy), and foot disease. Evidence suggests that the frequency with which these tests are performed in Thailand is suboptimal. A study in urban primary care clinics found that annual eye and foot examination was performed only in 21.5% and 45% of patients, respectively [15] while the ADA recommends annual testing starting at diagnosis [48].
Another study in the outpatient department of a university hospital found that the annual eye, urine albumin excretion, serum creatinine, and foot examination was performed in 38.4%, 42%, 83.5%, and 17.3% of patients in 2006 [49].
The highest rate for DMT2 patients receiving retinal examination, 75.6% between April to December 2003, was found in the diabetes registry project which included tertiary diabetes clinics [50].

Diabetes complications and cost of complications
Overall, prevalence of diabetic retinopathy (DR) ranged between 13.6-31.2% and it mostly involved non-proliferative DR while the prevalence of diabetes nephropathy ranged between 24-43.8% and was higher in patients with concomitant DR (Table 3). A variety of vascular complications ranging from absence of foot pulse to ulcer and gangrene leading to foot amputation and stroke were reported. Two studies reported the prevalence of chronic kidney diseases (CKD). A study in a primary care diabetes centre reported a prevalence of 25-27 (depending on the estimation method) for CKD stage 3-5 in 2007 [51] while a second study in the outpatient department of a tertiary hospital reported a higher prevalence (37% stage 3, 8.2 stage 4, 2.7 stage 5) in 2006 (Table 3) [49].
Complications have a major impact on the cost of diabetes (Table 3). A study predicted the cost of diabetes to increase up to 232% depending on the type of complication [52].
The median cost of illness for patients with complications was significantly higher than for people without complications (USD$ 479.93 vs. USD$ 115.12, p < 0.001) and increased with increasing numbers of complications (p < 0.001) [38].
Disability was a major driver of diabetes cost as well as complications due to vascular problems leading to heart failure and corresponding surgery.
In terms of potential savings from better prevention, a study showed that preventing gangrene in DMT2 patients would generate almost USD$ 250 (THB 10,000) per patient per year [52].

Diabetes outcomes
There are no national diabetes outcome indicators routinely collected in Thailand. Some diabetes centres report outcomes but this practice is voluntary and not standardised across the country.
The percentage of diabetes patients who were treated and controlled increased between 2004 and 2009 yet it remained low in addition to showing a large gender gap (men 7.7% in 2004 and 17.5% in 2009, women 15.8% in 2004 and 33.9% in 2009) [11]. Treatments rates for diagnosed patients were high leaving only 5.6% men and 1.9% diagnosed women without treatment [11]. However, 30% of treated men and 41% of treated women still did not attain diabetes control [11].
Diabetes control in patients with high blood pressure and high total cholesterol improved in 2009 from 3.4% to 12.2% in men and from 6.4% to 13.8% in women for hypertension between 2004 and 2009, 4% to 16.3% in men and 3.8% to 17.3% in women for high cholesterol between 2004 and 2009 [11]. However, at the same time the proportion of treated but not controlled patients increased from 25.4% to 36.3% in men and from 31.3% to 54.9% in women for hypertension and from 7.6% to 12.6% in men and from 7.0% to 18.5% in women for high cholesterol [11].
Evidence from tertiary care units in earlier studies (2003) shows a very high proportion of diabetes patients with poor glycaemic control (more than 70% of diabetes patients with HbA1c >7%) (

Access to treatment and inequalities
In principle, availability of medicines to treat diabetes in Thailand should be adequate since the universal health coverage policy covers more than 75% of the Thai population (the rest of the population is covered either by the civil servant or the social security scheme). Essential diabetes drugs such as metformin, sulfonylurea, pioglitazone, and insulin are included in the national drug list to which all UC insurees have access to. However, a study on diabetes mortality found that patients on the UC scheme were more likely to die than patients on the civil service scheme (adjusted hazard ratio 1.96, 95% CI 1.48-2.58, p-value < .005) [54]. Possible reasons for this include lower level of education and socioeconomic status, reduced access to lipid-lowering treatment (at the time of the study) and renal replacement among UC insured people in comparison to civil servants [54,55]. When the study was conducted between 2003 and 2006, low-cost generic statins were not yet available. This meant that many patients were not able to access treatment because of the high cost of the original drug. However, since the introduction of generic statins and their inclusion in the UC benefit package, this is no longer an issue.
Inequalities in access to treatment still persist in Thailand despite universal health coverage. Renal replacement therapy (RRT) for example was initially excluded from the benefit package of the UC scheme because the annual cost of haemodyalysis was four times higher (BHT 400,000, USD 12,100) than the price per quality-adjusted life year threshold set by the National Health Security Office (BHT 100,000, USD 3,0000) [9]. In 2008, thanks to pressure from patients and the public, renal replacement therapy was finally included in the UC benefit package. However, patients opting for haemodyialysis (which is more expensive than peritoneal dialysis) need to pay one third of the treatment cost out-of-pocket which corresponds to an annual cost of (BHT 133,333, USD$ 4,033) [9] in a country with a per capita gross domestic product of USD$ 4,608 in 2010 [1]. Further, geographical barriers can affect access to treatment and monitoring for patients living in remote areas.

Policies
In 2004, the Ministry of Public Health launched the programme "Healthy Thailand" in an attempt to address the growing burden of NCDs. This programme aimed to promote a healthy lifestyle and to screen 60% of population aged over 40 years to be screened for IFG and diabetes by the end of 2006 [12]. However, due to lack of data, it is not clear whether this target was achieved.
Reduction of diabetes morbidity and mortality rates due to diabetes is one of the 17 targets included in the Health Thailand Strategy 2004-2015 [56]. However, this target does not include measurable objectives and a strategy to achieve this reduction.
Mobile eye care is the new project that headed by the Ministry of Public Health to improve the access to diabetes care with the aim of preventing blindness from diabetes retinopathy. People living in rural areas often face difficulties in accessing health care. They often have to travel long distances and the transportation costs can be unaffordable to many. One way of alleviating this is the use of mobile clinics which travel to remote areas and provide health care to where people live. The project will be initially implemented in 11 provinces in the northern and north-eastern of Thailand. With time, the plan is to expand the project to cover the entire country [57].
The Thailand healthy lifestyle strategic plan (2011-2020) lists reduction of diabetes incidence, complications, disability, mortality, and expense as one of its five main development goals together with reduction of hypertension, heart disease, cerebrovascular disease and cancer [39]. This should be achieved through the promotion of a balanced diet, adequate physical exercise, and suitable emotional management. However, none of the eighteen short-to long-term performance indicators is linked to measurable objectives in terms of disease burden and cost reduction. Instead they simply mandate for either a reduction or an increase of the relevant indicator. At a higher level, the Thai healthy life-style strategy aims to establish political will, raise public awareness, focus on preventive measures, seek wider collaboration and strengthening active engagement from all stakeholders including public and private sectors, civil society and the general public. Finally, it also seeks to institutionalise organizational structures at all levels of the society, from national to village level, to serve as implementation units responsible for coordination, policy direction, budget allocation, and monitoring and evaluation [39].

Challenges in diabetes management according to Thai diabetes experts and senior public health officials
The strong features of the national policy and implementation framework for diabetes prevention and control in Thailand include the existence of capable staff and competent healthcare workers throughout Thailand, the presence of academic experts connected nationally and internationally, effective senior level management, strong policy development processes at the national level, and the experience of successfully implementing the tobacco control programme. Weaknesses include unorganised local and mid-level management, staff shortages, high workloads in rural areas, little time or opportunity for continuing training, and weak resource management.
Key opportunities would focus on a strong national policy response to diabetes and other NCDs, a strong network of competent health care workers and hospitals, budget support from both public and private sectors, and scientific and technical support from academics and researchers. The main threats identified are the negative influence of the media leading to changes in lifestyles and increasing risks for diabetes; inconsistent and unreliable information about diabetes disseminated to the public; low public awareness of diabetes issues especially among less well educated people; and high healthcare worker turnover in rural areas.

Prevalence
The NHES is a nationally representative survey of the health status of the Thai population. NHES III and IV found a higher prevalence of DMT2 in women, older individuals, and in urban areas b . Under-diagnosis was higher in men (2009: 47.3 vs. 23.4%, respectively; P < 0.001) [11,12] and in those with less than secondary education [11]. Both surveys showed that nearly all patients who were diagnosed with diabetes were also treated with glucose-lowering medications (2004: men 2.6%, women 1.7%; 2009: men 5.6%, women 1.9% of diagnosed patients were treated) [11]. However, despite the high treatment rates, the percentage of treated and controlled patients (receiving treatment with glucose-lowering medication and with FPG <7.2 mmol/L) was still low (2004: men 7.7%, women 15.8%; 2009: men 17.5%, women 33.9% treated and controlled patients) [11].
The high percentage of treated patients among diagnosed diabetics suggests the availability of a resilient health care system in terms of access to treatment. However, access is hampered by the low diagnosis rates which need to be improved by increased screening of high-risk groups. A risk score to identify individuals at high-risk of developing diabetes was developed for the Thai population [33]. Wider implementation of this lowcost instrument can help identify high-risk individuals to be screened and therefore to increase the percentage of diabetes patients who is diagnosed. This would allow earlier start of treatment and could help in preventing part of the costs of complications arising from neglect of the disease. Another issue is the low rates of treated and controlled patients. To address these issues evidence is needed on factors responsible for poor treatment   [22,25], this means that in addition to not capturing diabetes type 1 cases who did not attend a hospital, the results were also importantly influenced by missing data from hospitals that did not participate in the survey. Under-diagnosis is likely to have played a major role in the 1990s and first half of the 2000s because of incomplete insurance coverage which created barriers to health care access. Furthermore, apart from the North-East region, no data is available after 1997 and the most recent incidence data for the North-East region are for 2005.

Incidence of type 2 diabetes in adults
The latest data on the incidence of DMT2 in Thai adults show an incidence rate of 13.6 in men per 1000 PY and 6.4 per 1000 PY in women [18]. However this data is out-of-date as it refers to the time period 2001 to 2005 and is not representative of the entire country as it based on urban high-socio-economic status individuals working in the healthcare sector.

Gestational diabetes
Due to different gestational weeks and age of the prospective mother when the glucose test was conducted, it is not possible to draw definitive conclusions on the evolution of the prevalence of GDM over time. Data only included women attending antenatal care which might be biased towards higher socio-economic groups due to the lack of universal coverage at the time of the surveys. Despite these challenges, two main findings seem to emerge. First, a very large difference (eleven-fold) was reported when using different diagnostic criteria (NDDG (1.4%) vs. WHO criteria (15.7%)) on the same sample of women and another recent study that used the new criteria of international association of the diabetes and pregnancy study groups (IADPSG) found the prevalence of GDM in Bangkok was 23.0% [58]. Second, use of eligibility criteria which limit screening to women at high risk of developing GDM independently of their age have shown to miss more than 50% of cases among women aged 30 and over.

Costs
Data on the cost of diabetes without complications mainly comes from a study in one-hospital in North-East Thailand. Generalisability of local studies on cost of illness is affected by variation of input prices across the country, the level of care of the hospital analysed, and the patient status. One study in Thailand for example showed that a visit to the regional hospital was 3.48 times more expensive than a visit to a community hospital (THB 1,181 vs. THB 339 in 2002) [59]. Another issue was the uncertainty and large variation around the results (very large standard deviations were reported).

Complications and cost of complications
According to data from the Thailand diabetes registry, diabetic nephropathy was the most common complication accounting for 43.9% of all complications followed by retinopathy (30.7%), ischaemic heart disease (8.1%), and cerebrovascular disease (4.4%) [13]. Another study in the out-patient department of a university hospital found a lower prevalence for diabetic nephropathy, 37%, a similar prevalence for retinopathy, 31.2% but a substantially higher prevalence for cardiovascular and cerebrovascular disease, 28.9% and 10.6%, respectively. The first study also highlighted the existence of high prevalence of risk factors for diabetes and its complications (dyslipidaemia, hypertension, and obesity (BMI ≥ 25 kg/m 2 ) was 73.3%, 63.3%, and 52.6% respectively) [13].
Complications are the single largest driver of the cost of diabetes because they require more intensive care such as hospitalisation and often surgery. Studies showed that the median cost of illness for patients with complications in comparison to patients without complications was USD$ 480 vs. USD 115 [37], USD$ 190 for diabetic foot, USD$ 260 for cerebrovascular event, USD$ 336 for gangrene in comparison to USD$ 101 for patients without complications [52]. Interestingly, the latter study which was based on retrospective data analysis and cost forecast,  found that the highest proportion of treatment cost was caused by pharmacy services (45%) followed by outpatient (24%) and inpatient services (16%) [52]. The first study found a similar share of inpatient costs (11%) but in addition to direct costs it also estimated indirect costs and the contribution of informal care, mortality, and permanent disability to the cost of illness (27.8%, 17.5%, 18.7%, respectively) [37] (Table 5). This data shed some light on substantial indirect costs caused by morbidity and reduced ability to work. Preventing complications, and related disability, by improving diabetes control is therefore of paramount importance to reduce the health and economic burden of diabetes. There is a large variation in the estimates of the prevalence and costs of complications in Thailand. More studies at different levels of care and covering all the regions are needed to get a full picture of the prevalence and the cost of diabetes complications in the country.

Conclusions
Based on the results of this study, the following priorities for the future management of diabetes in Thailand were identified. First, increasing screening of diabetes in high risk population and promoting annual screening of diabetes complications in all diabetic patients. Second, identifying and addressing factors affecting poor treatment outcomes in light of reducing the number of treated but not controlled patients and therefore reducing their likelihood of developing complications. Third, policy should specify clear targets and provide and use a monitoring framework to track progress.
Fourth, efforts are needed to further improve data availability. Up-to-date data on the medical and economic burden of diabetes representative at the national level and at least the regional level are essential to identify needs and monitor progress towards established targets. Priority areas for data collection include incidence of diabetes in children and adults, prevalence of GDM, cost of diabetes and its complications, and treatment compliance and outcomes at individual level. Availability of prevalence data is good due to the regular NHES, however, these surveys are only conducted every 5-7 years.
Fifth, promotion of a healthy lifestyle for prevention of diabetes through education and quality information delivered to the public.
Efforts to address these issues have already started in some areas of the country but not overall. In order to achieve this, a multisectoral effort including concerted policy actions from a variety of policy makers (beyond the ministry of public health and including other relevant ministries) and of public opinion leaders as well as interventions involving public and private delivery channels is required.

Endnotes
a Study year is not specified but the publication data is 2000 so it is very likely to refer to the late 1990s.
b In 2004 prevalence in urban areas was higher for men but not for women.