The present study is a nationwide, multicenter, cross-sectional survey of T2DM outpatients across Thailand. All data were obtained from the Diabetes and Hypertension dataset, an ongoing nationwide project, titled: "An assessment on quality of care among patients diagnosed with type 2 diabetes and hypertension visiting hospitals of ministry of public health and Bangkok metropolitan administration in Thailand, 2010-2012." This project is administered by the Medical Research Network of the Consortium of Thai Medical Schools (MedResNet) Thailand, under the sponsorship of the Thai National Health Security Office. Patients and hospitals were sampled using a proportional to size stratified cluster sampling approach, and then T2DM outpatients were proportionally sampled from the 602 participating hospitals across Thailand during the period April 1st to June 30th for both years 2011 to 2012.

A two stage stratified cluster, proportional to size sampling approach was used to select a nationally and provincially representative sample of diabetes patients in Thailand. The first stage was ensure that the number and levels of hospitals were appropriately sampled across provinces to represent the diabetes population in Thailand. The second stage of cluster sampling was to ensure that the right mix of hospitals (levels) were sampled (given attendance patterns in Thai diabetes patients). Thailand has three levels of hospitals: regional hospitals (>500 beds), general (provincial; 200 to 500 beds) hospitals, and community hospitals, with this last group being further subdivided into large (80 to 120 beds), medium (60 beds), and small (10 to 30 beds) community hospitals. All regional and general hospitals in Thailand were included in our study, but only 456 (61.96% of the total 736) appropriately sized community hospitals were included. The 456 hospitals were made up of 10%, 20%, and 70% of large, medium, and small community hospitals, respectively, reflecting the size distribution of these hospitals throughout the community. Finally, once the list of participating hospitals was compiled, proportion to size sampling was performed to ensure the appropriate proportion of diabetes patients (based on national attendance patterns across hospital levels) were sampled at the five different levels of hospitals.

Patient information was retrospectively collected via medical records. Diabetes and Hypertension dataset, study protocols and case report forms are archived at the DAMUS (Data Archival for Maximum Utilization System) website [20]. The study was approved by the Ethical Review Committee for Research in Human Subjects Ministry of Public Health, Thailand, and the Khon Kaen University Ethics Committee for Human Research, Thailand. Informed consent was obtained from all patients.

Patients with T2DM and receiving medical care in participating hospitals for at least 12 months were included in the present study. Patients with incomplete information in the key study variables (outcome and hypertension status) were excluded. After evaluating by eligibility, a total sample of 55,797 T2DM patients was considered (Fig. 1).

The outcome variable in the present study is microvascular complication diagnosis (+/-) in the last 12 months, which includes any of diabetic neuropathy, diabetic nephropathy, diabetic retinopathy, and renal insufficiency diagnosis (+/-) in the last 12 months. All microvascular complications were documented by medical records. Diabetic neuropathy was defined as the presence of symptoms and/or signs of peripheral nerve dysfunction in people with diabetes after the exclusion of other causes. Diabetic nephropathy was defined as serum creatinine level equal or greater than 2 mg/dL, or positive dipstick proteinuria, or the presence of random micro albuminuria/creatinine ratio greater than 30 mg/g. Diabetic retinopathy was diagnosed by detailed fundus examination, including either proliferative or background diabetic retinopathy diagnosed by an ophthalmologist. Renal insufficiency was defined by the reduction in estimated glomerular filtration rate (GFR): an estimated halving of GFR, and/or a 25 mL/min/1.73 m² decline in GFR from baseline.

Hypertension comorbidity (+/-) is the main study effect of interest. Hypertension was defined as blood pressure exceeding 140 over 90 mm Hg (a systolic blood pressure above 140 mm Hg or a diastolic blood pressure above 90 mm Hg) [21], as previously diagnosed by a physician. Other continuous covariates considered include: (1) age, (2) diabetes duration, and clinical parameter during the last 12 months follow-up: (3) fasting plasma glucose (mmol/L); (4) serum creatinine (mmol/L); (5) total cholesterol (mmol/L); (6) triglyceride (mmol/L); and (7) high density lipoprotein cholesterol (HDL-C mmol/L). Categorical covariates considered include: (1) glycosylated hemoglobin (HbA1c) treatment target (<7%, ≥7%); (2) blood pressure treatment target (blood pressure: <130/80, ≥130, or ≥80 mm Hg); (3) low density lipoprotein cholesterol treatment target (LDL-C; <100, ≥100 mg/dL; based on the American Diabetes Association 2012 guideline which is currently used in the Thai diabetes care system [22]); (4) previous macrovascular complication (diagnosis >12 months ago); and (5) previous microvascular complication (diagnosis >12 months ago); (6) diabetes medication (no treatment, oral hypoglycemic agents [OHAs], insulin, both OHAs, and insulin) and finally other covariates such as: health care coverage type, hospital type (regional, provincial, and community), smoking status (never, previous, and ongoing smoking), gender, and body mass index (BMI) class (classified <18.5, 18.5 to 22.9, 23 to 27.4, 27.5 to 32.4, 32.5 to 37.4, ≥37.5 kg/m² [23]).

Frequencies, means and standard deviations, were used to described categorical and continuous variables, respectively. The effects of the hypertension comorbidity, along with other risk factors on the presence of macrovascular complication were investigated using binary logistic mixed effect regression. The potential clustering effect introduced by hospital and the multi-level nature of the data (hospital type is an upper level effect) were accounted for by using a mixed effect modeling approach. We investigated the potential bias introduced by missing values in the dataset using multiple iterative regression imputation. By comparing the difference between the magnitude of effects in the complete case and imputed data analyses, missing values bias could be gauged. Pooled estimates were generated across five imputed datasets and the global likelihood ratio test (was) run using the methods outlined by Meng and Rubin [24] (1992). Purposeful selection of covariates was used to build the best model for the presence of microvascular complication [25]. Crude and adjusted odd ratios with 95% confidence interval (CI) and P value were used to represent the association between hypertension and other covariates with microvascular complication. All data analyses were conducted using the R statistical programming language (v.3.0.3) [26], the R library lme4 [27] was used for the mixed modeling, and the R library mi [28] was used for the multiple imputation.

The study flow is shown in Fig. 1. Of the 55,797 T2DM patients included in our sample, 55.35% were hypertensive. Table 1 provides the main characteristics of the study sample. Prevalence of microvascular complication diagnosis in the last 12 months was higher in T2DM patients with hypertension than those without hypertension (12.12% vs. 9.80%, respectively). The mean age, diabetes duration, and serum creatinine in T2DM with hypertension were higher than those without hypertension, while mean total cholesterol, triglyceride, and HDL-C level were comparable between the two groups (Table 1). Prevalence of patients achieving HbA1c and LDL-C treatment target in hypertensive patients (36.31% and 44.13%, respectively) was considerably higher than non-hypertensive patients (28.36% and 40.94%, respectively). However, hypertensive patients tended to achieve the blood pressure treatment target less often compared to non-hypertensive patients (32.76% vs. 52.44%). Prevalence of both previous macrovascular and previous microvascular complications in hypertensive patients were higher than non-hypertensive patients. Perusal of Table 1 shows minimal difference between available cases and imputed data.

Table 2 gives the number and proportion of patients in various blood pressure classes and a prevalence estimate of microvascular complications for each of these groups. Perusal of Table 2 indicates that the prevalence of microvascular complications remains relatively stable from the normal blood pressure group (12.58%; 95% CI, 12.08 to 13.04) through to prehypertension group (12.93%; 95% CI, 12.30 to 13.59). However, the prevalence of microvascular complications was significantly higher for the stage 1 hypertension group (14.57%; 95% CI, 14.0 to 15.17). It is also likely that the prevalence is higher in the stage 2 hypertension patients, but the relatively small number of patients in this class leads to a wide 95% CI (16.51%; 95% CI, 12.71 to 21.13).

The crude and adjusted estimates of the risk factors for microvascular complication are shown in Table 3. After adjusting for other covariates using multivariable binary mixed effect modeling, patient with the hypertension comorbidity had 1.32 time the odds of developing microvascular complication (adjusted odds ratio [OR_adj], 1.32; 95% CI, 1.20 to 1.46; P<0.001).

In the final multivariable model, older T2DM patients and those with longer diabetes duration had 1.07 and 1.21 times the odds of developing microvascular complication, per 10 years (age) and 5 years (duration), respectively (OR_age, 1.07; 95% CI, 1.03 to 1.12; P<0.01; and OR_duration, 1.12; 95% CI, 1.07 to 1.16; P<0.001, respectively). Diabetes medication types were shown to be significantly associated with an increased microvascular complication. The odds of patients with insulin treatment and insulin with OHAs treatment were 2.71 and 1.83 higher to have microvascular complication compared to no medication (OR_insulin, 2.71; 95% CI, 2.02 to 3.64; and OR_insulin with OHAs, 1.83; 95% CI, 1.37 to 2.44, respectively).

The multivariable analysis also demonstrates that higher BMI class, lower hospital level, achievement of HbA1c and blood pressure treatment target, higher HDL-C, and previous microvascular complication were all shown to be associated with microvascular complication among patients with T2DM. For instance, previous microvascular complication were shown to be significantly associated with developing microvascular complications, with the odds of developing microvascular complications decreasing by 58% (OR_adj, 0.42; 95% CI, 0.37 to 0.48; P<0.001).

Fig. 2 shows a comparison of the OR_adj and CIs results derived from the complete case and imputed data analyses. The results indicated minimal bias was introduced by missing values, and importantly, the significance of the hypertension effect to microvascular complication among patients with T2DM did not change. Fig. 2 does indicate that there were some changes in the significance of some other effects. Importantly, the underweight BMI level effect could not be shown to be significant in the imputed data analysis.