Factors associated with type 2 diabetes in patients with vascular dementia: a population-based cross-sectional study

The Department of Health in Taiwan implemented the National Health Insurance (NHI) program in 1995. By the end of 1996, approximately 96% of all residents in Taiwan had enrolled [8]. The coverage rate increased steadily from 96.1 to 98.6% from 2000 to 2007. All hospitals and clinics contracted with the NHI program are required to submit patient claims to receive reimbursements from the program. To verify the accuracy of claims data, the Bureau of NHI Management performs quarterly expert reviews on a random sample of inpatient claims at each hospital and clinic. Severe penalties are issued for false reports. We linked data from Taiwan’s Registry for Catastrophic Illness Database and related outpatient and inpatient claims datasets from 1997 to 2008.

For this population-based cross-sectional study, we included all patients in Taiwan with dementia who were newly registered to receive the catastrophic illness certification for senile dementia according to the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM; code 290.X) from 1997 to 2008. These designations are based on rigid diagnostic criteria evaluated by a neurologist or psychiatrist. The registry date found on the catastrophic illness certification was considered the index date for this study. All patients were defined as having comorbid diabetes if diagnostic ICD-9-CM code 250.XX appeared on at least two ambulatory care claims records or at least one inpatient care claims record within 1 year leading up to dementia diagnosis.

Several factors were considered to assess their potential associations with type 2 DM, namely demographic factors (age, sex, area of residence, urbanization level, and insurance amount), comorbidities (myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, peptic ulcer disease, mild liver disease, renal disease, cancer, hypertension, and hyperlipidemia), and disease severity. Area of residence was classified into four regions of Taiwan (north central, south, and east); urbanization level was divided into urban and rural; and insurance amount was divided into dependent, < NT$20,000 per month, and ≥ NT$20,000 per month. Patients were defined as “low income” when their insurance amount was dependent, “medium income” when their insurance amount was < NT$20,000 per month, and “high income” when their insurance amount was ≥ NT$20,000 per month. Comorbidities were defined based on ICD-9-CM codes (Additional file 1: Table S1) listed on two or more ambulatory care claims records or one or more inpatient care claims records during the year prior to dementia diagnosis. To further evaluate the associations of various systematic diseases with diabetes, we grouped these comorbidities into four main categories based on similar manageable risk factors. These categories were (1) cardiovascular and cerebrovascular diseases (myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, and hypertension), (2) digestive system diseases (peptic ulcer disease and mild liver disease), (3) renal and metabolic system diseases (renal disease and hyperlipidemia), and (4) cancer. Disease severity was defined based on Charlson comorbidity index scores, which are listed in a previous report [9], calculated based on diseases.

Participant characteristics were analyzed as follows. Distributions of continuous variables were expressed as mean ± standard deviation (SD) or median (interquartile range), and those of categorical variables were expressed as numbers and percentages. The differences in the distributions of count variables between DM and non-DM were analyzed through Mann–Whitney U testing, and those of categorical variables were analyzed through the chi-squared (χ²) testing. Multivariable logistic regressions adjusted for all demographic factors, and comorbidities or various systematic diseases were tested to identify independent factors associated with DM. In addition, we investigated associations among various systematic disease groups, the number of each systematic disease, and DM by using multivariable logistic regression adjusted for patient characteristics. Data were represented as odds ratios (ORs) and 95% confidence intervals (CIs). To further explore the effects of income and comorbidities on the prevalence of DM, the proportion of DM and high income by number of comorbidities, and the proportions of various systematic diseases by DM and income were also investigated. To validate our main findings, we conducted sensitivity analysis after redefining the DM group as ICD-9-CM code 250.XX appearing on at least three ambulatory care claims records or at least one inpatient care claims record within 1 year leading up to dementia diagnosis. All statistical operations were performed in SAS (version 9.4, SAS Institute, Cary, NC, USA). A p value of < 0.05 was considered significant.

In the study cohort, 22.5% patients with dementia also had diabetes. Table 1 provides a summary of patient characteristics. Compared with patients without diabetes, those with diabetes were significantly more likely to be female, young, and living in south and rural areas (Table 1). They were also more likely to have low incomes, more comorbidities (myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, peptic ulcer disease, mild liver disease, renal disease, cancer), and a greater number of severe diseases. The top three largest differences in the distributions of comorbidities between the DM and non-DM groups were cerebrovascular disease (percentage difference: 18.9), peptic ulcer disease (percentage difference: 7.1), and renal disease (percentage difference: 5.7).

Table 2 shows the associations of baseline characteristics and comorbidities with DM. Age, sex, area of residence, urbanization level, insurance amount, congestive heart failure, cerebrovascular disease, peptic ulcer disease, mild liver disease, renal disease, hypertension, and hyperlipidemia were significantly and independently associated with DM in patients with dementia (Table 2). Notably, patients with DM were more likely to be female (adjusted OR: 1.44, 95% CI: 1.36–1.52) and young (range of adjusted OR: 0.55–1.13). Those with DM were significantly more likely to have more comorbidities, have lower incomes, and live in rural areas. We analyzed the associations of DM with income, number of comorbidities, and systematic diseases (Figs. 1 and 2). As can be seen in Fig. 1, which shows the results of our analysis regarding the prevalence of DM, number of comorbidities, income, the prevalence of DM was higher in patients with four or more comorbidities and lower in those with high incomes (Fig. 1). Figure 2 depicts the results of our analysis regarding the proportion of comorbidity-related diseases based on diabetes status and insurance amount. Patients with DM had substantially more comorbidities than did those without DM. Those with DM and those in the lower income group had the highest proportion of renal and metabolic system diseases (Fig. 2). The sensitivity analysis strongly supported these model findings (Additional file 1: Table S2).

Table 3 shows the results of our study regarding association between systematic comorbidities and DM in patients with dementia. Prevalence of DM was significantly associated with the following systematic comorbidities: cardiovascular diseases (adjusted OR: 3.93, 95% CI: 3.68–4.19); digestive system diseases (adjusted OR: 1.34, 95% CI: 1.26–1.42); renal and metabolic system diseases (adjusted OR: 2.81, 95% CI: 2.64–2.98); and cancer (adjusted OR: 1.23, 95% CI: 1.09–1.38) (Table 3). The likelihood of DM was clearly higher in patients with one or two renal or metabolic system diseases than in those without (adjusted OR: 3.05, 95% CI: 2.86–3.25 and adjusted OR: 5.32, 95% CI: 4.18–56.78, respectively). Similar results from the sensitivity analysis are shown in Additional file 1: Table S3.