Type 2 diabetes mellitus (DM) is a chronic disease caused by insulin resistance and progressive insulin secretion defects, which develop over time, resulting in high levels of sugar in the blood (Chatterjee et al. 2017). In 2018, the prevalence of all types of diabetes mellitus among the elderly (aged over 65 years) in Korea was 27.6%. Although Korean diabetic patients showed high disease recognition (77.9%) and treatment rates (72.9%), their blood sugar control rate was generally low (29.3%) (Jung et al. 2021). Poor control of blood sugar levels causes severe complications in the circulatory, kidney, and nervous systems. The mortality of diabetic patients is due more to diabetic complications than the disease itself, and chronic conditions, such as microvascular and macrovascular complications, have a major impact on long-term prognosis (Nwaneri et al. 2013). The risk of cardiovascular disease is two to four times greater in type 2 DM patients than in non-diabetic patients, and cardiovascular complications are the cause of death in 65-70% of the cases among diabetic patients (Lim et al. 2009). In addition, diabetic patients are 20 times more likely to be hospitalized because of non-traumatic leg amputation than non-diabetic patients (Public Health Agency of Canada 2011).

Dementia is a collection of symptoms that result from an ongoing decline in brain function. The major symptoms of dementia include problems with memory, thinking speed, and understanding, and difficulties in performing daily activities (Arvanitakis et al. 2019). With the aging of the population, the number of dementia patients has increased. The prevalence of dementia among people aged over 65 in Korea is expected to increase to 1.75 million cases by 2050 (National Institute of Dementia 2020). A diabetes-dementia complex disease, called ‘type 3 diabetes’, has emerged as an important public health issue (de la Monte and Wands 2008). Diabetic patients with dementia have greater difficulty in managing their blood sugar levels than patients without dementia. Those with dementia have difficulties explaining their symptoms and remembering how they were instructed to manage their blood sugar levels. In addition, because of these difficulties, poor compliance with anti-diabetic medications can be expected. It is virtually impossible for dementia patients to manage their blood sugar levels by themselves; thus, support from guardians or caregivers is necessary (Yoo 2012). Thus, it is assumed that patients with diabetes and dementia have a relatively higher risk of diabetic complications than those without dementia.

Accordingly, we conducted an empirical study to assess whether the presence of dementia is associated with an increased risk of diabetic complications in elderly Korean patients with type 2 DM, using a real-world dataset from the national insurance claim records. We expect that our study results will help emphasize the need for and importance of adequate support for diabetic patients with dementia; in particular, help in blood sugar management to prevent diabetic complications.

Study patients and data source

We used the 2018 Health Insurance Review and Assessment Service-Aged Patient Sample data (2018-HIRA-APS, serial number: HIRA-APS-2018-0043), which are nationally representative cross-sectional data provided by the government agency Health Insurance Review and Assessment Service (HIRA). The HIRA-APS data provide claim records of a 10% random sample (approximately 700,000 patients) of Korean elderly patients (over 65 years old) enrolled in the National Health Insurance (NHI) or Medical Aid (MA) each year. The NHI program is a wage-based, mandatory insurance program covering approximately 96% of the population, and the MA program is a government-subsidized public assistance program for poor and medically indigent individuals (Kim et al. 2022). The study protocol was approved by the Institutional Review Board of Yonsei University, Seoul, South Korea (IRB No: 7001988-202208-BR-1653-01E). The requirement for informed consent of the study subjects was waived by the board.

Patients with type 2 DM were defined as those with at least one claim record of all diagnosis codes of type 2 DM (International Classification of Diseases codes, 10th revision (ICD-10): E11) in 2018 (Safieddine et al. 2021). If patients with type 2 DM had at least one claim record with all diagnosis codes of dementia (ICD-10: F00, F01, F02, F03, or G30) in 2018 (Bauer et al. 2014), we defined them as “type 2 DM patients with dementia” or “dementia patients with type 2 DM”; otherwise, they were defined as type 2 DM patients without dementia. Because our data are one-year cross-sectional data, it is difficult to identify the temporal order—that is, which disease, type 2 DM or dementia—occurred first in the patient. Nevertheless, we conceptually denote those with claim records showing diagnoses of both diseases either simultaneously or separately during the study period as “type 2 DM patients with dementia.” Among 224,984 patients with type 2 DM from the 2018 HIRA-APS data, we identified 4,569 type 2 DM patients with dementia and 220,415 type 2 DM patients without dementia.

We performed 1:4 propensity score (PS) matching between the two groups (i.e., type 2 DM patients with dementia and without dementia) using patient demographic characteristics, including gender, age, and the type of National Health Security (NHS) program in which they were enrolled (i.e., NHI or MA), to adjust for potential confounding effects of those variables. PS matching is a statistical method that produces balanced matched samples (Gu and Rosenbaum 1993). Thus, we can minimize selection bias and increase the validity of causal inference (Guo and Fraser 2010). In this study, caliper matching was performed based on PS using a greedy algorithm. To estimate the risk differences, a caliper width of 0.1440701 was used, which is equal to 0.2 of the standard deviation of the logit of the PS, to minimize the standardized mean difference (SMD) (Austin 2011). The standardized mean difference was calculated as follows:

$$SMD=\frac{\left(p_1-p_2\right)}{\sqrt{\frac{p_1\left(1-p_2\right)+p_2\left(1-p_1\right)}{2}}}$$

where p1 and p2 are the prevalence of dichotomous variables in the dementia and non-dementia groups, respectively (Zhang et al. 2019). In Fig. 1, we show the screening process to select the patient records included in the study.

Figure 1. Flow chart for identifying patients included in the study. n, number of patients; DM, Diabetes Mellitus; HIRA-APS, Health Insurance Review and Assessment‒Aged Patient Sample; ICD, International Classification code; PS, propensity score.

Complication of DM

We used two outcome variables to investigate whether the study participants experienced complications of DM during the study period. First, if the study subjects had at least one claim record with a diagnosis code for complications of DM, we considered them to have diabetes complications. Based on previous studies, diagnoses referring to DM complications and their ICD-10 codes were identified, as shown in Table 1 (Cho et al. 2016; Dugan and Shubrook 2017). Second, if the study participants had at least one claim record of hospitalization with a diagnosis code for complications of DM, we assumed that they had experienced severe diabetes complications.

Table 1 . List of complications of type 2 diabetes mellitus included in this study.

ICD-10 code	Name of diagnosis
E11.0	Type 2 diabetes mellitus with hyperosmolarity
E11.1	Type 2 diabetes mellitus with ketoacidosis
E11.2	Type 2 diabetes mellitus with kidney complications
E11.3	Type 2 diabetes mellitus with ophthalmic complications
E11.4	Type 2 diabetes mellitus with neurological complications
E11.5	Type 2 diabetes mellitus with circulatory complications
E11.6	Type 2 diabetes mellitus with other specified complications
E11.7	Type 2 diabetes mellitus with multiple complications
E11.8	Type 2 diabetes mellitus with unspecified complications

ICD-10: International Classification of Diseases, 10^th revision..

Data analysis

Odds ratios (ORs) were calculated to compare the risk of diabetic complications and hospitalization (with diabetic complications) between dementia and PS-matched non-dementia groups. We conducted multiple regression analyses to examine whether accompanying dementia is independently associated with the risk of diabetic complications and hospitalization with diabetic complications, respectively, after controlling for the confounding effect of other covariates. Because both dependent variables are binary, and because treatment subjects (i.e., dementia patients) were matched with non-treatment subjects (i.e., non-dementia subjects), we chose a conditional logistic regression model (Koletsi and Pandis 2017). The Charlson Comorbidity Index (CCI) was included as a covariate in the regression model to adjust for the potential effect of the severity of the patient’s baseline health status on the risk of diabetic complications. For individual study subjects, the CCI was computed for the remaining diseases after excluding type 2 DM (E11), type 2 DM complications (E11.0-E11.9), and dementia (F00, F01, F02, F03, and G30).

To examine whether there was a difference in the type of diabetic complications in the presence of comorbid dementia, we compared dementia and PS-matched non-dementia groups with respect to the distribution of the most frequent diagnoses of diabetic complications in the claims records. All statistical analyses were performed using the SAS software (version 9.4; SAS Institute, Inc., Cary, NC, USA).

Characteristics of the study subjects

This study included the records of 4,569 patients with type 2 DM and dementia and 220,415 patients with type 2 DM but without dementia. The demographic characteristics were compared between the two groups (Table 2). Before PS matching based on age, sex, and type of NHS program enrolled (i.e., NHI or MA), the dementia group comprised more females (64.39%) than did the non-dementia group (56.16%, p < 0.001). The proportion of patients aged ≥ 75 years in the dementia group (72.71%) was approximately twice that of the non-dementia group (44.12%, p < 0.001). The proportion of MA patients in the dementia group (12.74%) was approximately 1.5 times that in the non-dementia group (8.07%). After 1:4 PS matching, we could select 4,569 type 2 DM patients with dementia and 18,276 type 2 DM patients without dementia as the final study subjects. The average CCI score in the dementia group (2.05) was significantly higher than that in the PS-matched non-dementia group (1.76; p < 0.001).

Table 2 . Basic characteristics of type 2 DM patients with and without dementia.

	No. of patients (%)					Standardized difference of mean (%)
	With dementia	Without dementia (Before PS matchinga)	p-valueb	Without dementia (After 1:4 PS matchinga)	p-valuec
Total	4,569	220,415		18,276
Gender			< 0.0001		1.0
Male	1,627 (35.61)	96,626 (43.84)		6,508 (35.61)		0.000
Female	2,942 (64.39)	123,789 (56.16)		11,768 (64.39)		0.000
Age (years)			< 0.0001		1.0
65-69	426 (9.32)	65,938 (29.92)		1,704 (9.32)		0.000
70-74	821 (17.97)	57,232 (25.97)		3,284 (17.97)		0.000
≥ 75	3,322 (72.71)	97,245 (44.12)		13,288 (72.71)		0.000
NHS program enrolled			< 0.0001		1.0
NHI	3,982 (87.15)	202,353 (91.81)		15,928 (87.15)		0.000
MA	587 (12.85)	18,062 (8.19)		2,348 (12.85)		0.000
CCI
Mean (±SD)	2.06 (1.05)	1.71 (1.11)	< 0.001	1.76 (1.10)	< 0.001
0	489 (10.70)	39,391 (17.87)	< 0.001	3,063 (16.76)	< 0.001
1	931 (20.38)	58,456 (26.52)		4,675 (25.58)
2	970 (21.23)	49,337 (22.38)		4,121 (22.55)
≥ 3	2,179 (47.69)	73,231 (33.22)		6,417 (35.11)

^aType 2 Diabetes mellitus (DM) patients with dementia were matched with those without dementia by propensity score based on demographic characteristics (i.e., gender, age, and type of the National Health Security program enrolled)..

^bChi-square tests comparing patients with type 2 DM with dementia vs. without dementia before PS matching..

^cChi-square tests comparing patients with type 2 DM and dementia vs. patients without dementia after PS matching..

CCI, Charlson comorbidity index; DM, diabetes mellitus; MA, Medical Aid; NHI, National Health Insurance; NHS, national health security; SD, standard deviation; PS, propensity score..

Association between dementia and diabetic complications

In comparing dementia and non-dementia groups, the OR for the prevalence of diabetic complications after PS matching was 1.328 (95% confidence interval (CI): 1.244-1.417) (Table 3). The risk of severe diabetic complications, defined as cases of hospitalization with diabetic complications, was 1.371 times higher (95% CI: 1.240-1.516) in the dementia group than in the non-dementia group (Table 3). After controlling for the effect of patients’ baseline health status using CCI in the conditional logistic regression analyses, the adjusted OR for the risk of diabetic complications decreased to 1.248 (95% CI, 1.169-1.333), whereas it increased to 1.748 (95% CI, 1.575-1.939) for the risk of hospitalization with diabetic complications (Table 4).

Table 3 . Risk of diabetes complications in patients with comorbid dementia and type 2 DM.

Type 2 DM patients	Prevalence of DM complications				Hospitalization with DM complications
	No. patients (%)		OR (95% CI)		No. patients (%)		OR (95% CI)
	Yes	No			Yes	No
With dementia (n = 4,569)	2,484 (54.37)	2,085 (45.63)	Ref.		1,636 (47.29)	5,574 (77.31)	Ref.
Without dementiaa (n = 18,276)	8,642 (47.29)	9,634 (52.71)	1.328 (1.244-1.417)		764 (28.70)	1,898 (71.30)	1.371 (1.240-1.516)

^aPatients without dementia were matched with patients with dementia by propensity score based on demographic characteristics (i.e., gender, age, and the type of National Health Security program enrolled)..

CI, confidence interval; DM, diabetes mellitus; OR, odds ratio; Ref, Reference group; PS, propensity score..

Table 4 . Conditional logistic regression analysis of the association between comorbidity of dementia and the risk of DM complication.

	DM complication				Hospitalization with DM complication
	β-coefficient	SE	Adj. OR [95% CI]		β-coefficient	SE	Adj. OR [95% CI]
Dementia
Without dementia	Ref.	-	-		Ref.	-	-
With dementia	0.1108	0.0168	1.248 [1.169, 1.333]		0.2791	0.0265	1.748 [1.575, 1.939]
Charlson Comorbidity Index
0	Ref.	-	-		Ref.	-	-
1	0.0772	0.0237	1.253 [1.150, 1.365]		−0.3561	0.0523	0.985 [0.807, 1.202]
2	0.0304	0.0244	1.395 [1.278, 1.522]		0.0671	0.0494	1.314 [1.083, 1.595]
≥ 3	0.3491	0.0211	1.918 [1.770, 2.079]		0.7637	0.0376	3.017 [2.540, 3.584]

Each regression model was adjusted for demographic characteristics, such as age, gender, and the type of National Health Security program using a propensity score matching method..

CI, confidence interval; DM, diabetes mellitus; OR, odds ratio; Ref, reference group; SE, standard error..

Distribution of the type of diabetic complications

There was no difference in the distribution of the type of diabetic complications between dementia and PS-matched non-dementia groups in terms of the ranking and proportion of patients (Table 5). We identified the most frequent diagnoses in claims records with diabetic complications. For both dementia and PS-matched non-dementia groups, type 2 DM with unknown complication (ICD 10-code: E11.8) was ranked the highest, followed by neurological complications (E11.4), circulatory complications (E11.5), kidney complications (E11.2), multiple complications (E11.7), other specified complications (E11.6), eye complications (E11.3), coma (E11.0), and acidosis (E11.1). The ranks of frequent complications were the same, either based on all types of healthcare services (i.e., inpatient and outpatient services) or inpatient services only. For most of the nine complications, the proportion of patients with diabetic complications was higher in the dementia group than in the non-dementia group.

Table 5 . Distribution of frequent diagnoses in claim records with diabetic complications.

Rank	In claim records for all types of health care services (inpatient and outpatient services)						In claim records for inpatient services
	Type 2 DM patients with dementiaa (n = 4569)			PS-matched type 2 DM patients without dementiaa (n = 18276)			Type 2 DM patients with dementiaa (n = 4569)			PS-matched type 2 DM patients without dementiaa (n = 18276)
	ICD-10 codes (description)	No. patients (%)b		ICD-10 codes	No. patients (%)b		ICD-10 codes	No. patients (%)b		ICD-10 codes	No. patients (%)b
1	E11.8 (Type 2 DM with unknown complication)	1,114 (44.85)		E11.8	3,391 (39.24)		E11.8	421 (55.10)		E11.8	885 (54.10)
2	E11.4 (Type 2 DM with neurological complication)	849 (34.18)		E11.4	2,676 (30.97)		E11.4	265 (34.69)		E11.4	512 (31.30)
3	E11.5 (Type 2 DM with circulatory complication)	571 (22.99)		E11.5	2,079 (24.06)		E11.5	180 (23.56)		E11.5	348 (21.27)
4	E11.2 (Type 2 DM with kidney complication)	408 (16.43)		E11.2	1,456 (16.85)		E11.2	164 (21.47)		E11.2	346 (21.15)
5	E11.7 (Type 2 DM with multiple complication)	218 (8.78)		E11.7	737 (8.53)		E11.6	118 (15.45)		E11.6	226 (13.81)
6	E11.6 (Type 2 DM with other specified complication)	212 (8.53)		E11.6	593 (6.86)		E11.7	98 (12.83)		E11.7	206 (12.59)
7	E11.3 (Type 2 DM with eye complication)	100 (4.03)		E11.3	359 (4.15)		E11.3	36 (4.71)		E11.3	79 (4.83)
8	E11.0 (Type 2 DM with coma)	52 (2.09)		E11.0	128 (1.48)		E11.0	29 (3.80)		E11.0	42 (2.57)
9	E11.1 (Type 2 DM with acidosis)	12 (0.48)		E11.1	34 (0.39)		E11.1	4 (0.52)		E11.1	15 (0.92)

^aPatients without dementia were matched with patients with dementia by propensity score based on demographic characteristics (i.e., gender, age, and the type of National Health Security program enrolled). ^bDuplication (> 100%). DM, diabetes mellitus; ICD-10, International Classification of Disease, version 10; PS, propensity score.

Our results showed that type 2 DM patients with dementia and aged > 65 years had a significantly higher risk of diabetic complications and hospitalization with diabetic complications than those patients without dementia. These results suggest that patients with diabetes accompanied by dementia were more likely to experience both mild and severe diabetic complications. Due to decreased cognitive function, it is difficult for patients with dementia to have a balanced diet and accurate drug therapy, and management of blood sugar levels and night hypoglycemia becomes difficult. Glycemic variability causes acute diabetic complications, such as hypoglycemic coma and hyperosmotic non-ketonic coma, as well as chronic complications, such as diabetic neuropathy, hypertension, and diabetic retinopathy. Therefore, our results can be considered as evidence of the critical need to assist diabetes patients with comorbid dementia in the management of their glycemic levels.

Because target blood glucose levels change depending on the type of complications, patients with type 2 DM and dementia should set and practice customized blood glucose goals depending on their particular complications (DeFronzo et al. 2015). In addition, studies have shown that if glycated hemoglobin is maintained at more than 7% (the threshold of hyperglycemia), the risk of cerebrovascular dementia increases. Thus, blood sugar levels should be carefully managed with a focus on both hypoglycemia and hyperglycemia (Zimmerman and Desemone 2010; Yaffe et al. 2012).

A frequency analysis of diabetic complications was performed to determine whether the risk of certain diabetic complications increased in the presence of dementia. Unknown complications occurred most frequently in both patient groups, followed by neurological complications, and the other frequency rankings were the same. These results can be interpreted as evidence that the types of complications that occur are independent of the onset of dementia, and dementia does not increase the risk of any specific diabetic complications.

This study had several limitations. Kidney and circulatory diseases are both major diabetic complications as well as causes of diabetes. Due to the limitation of one-year cross-sectional data, we were not able to determine the temporal association between the onset of type 2 DM and these diseases. Thus, we included only the kidney (ICD-10 codes: E11.2) and circulatory diseases (E11.5), denoted as diabetic complications. This may have led to the under-identification of diabetic complications in this study. Second, we defined those with claim records of type 2 DM and dementia as “type 2 DM patients with dementia.” However, in cross-sectional data, it is not clear which of the diseases—type 2 DM or dementia—occurred beforehand. Thus, with respect to the datasets used in this study, patients with type 2 DM and dementia can be considered as dementia patients with type 2 DM. However, which of the two diseases was a main or comorbid condition did not seem to affect the results of our examination of the association between the presence of dementia and the risk of diabetic complications.

In summary, diabetic patients with dementia had a significantly higher risk of diabetic complications, regardless of severity, than those without dementia. Given that dementia can occur as a neurological complication of diabetes, diabetic patients with dementia are not rare among elderly diabetic patients. When dementia occurs in patients with diabetes, it is very important to thoroughly manage blood sugar levels to prevent additional diabetic complications. However, due to decreased cognitive function and memory, dementia patients have difficulty monitoring glycemic levels regularly and controlling glycemic levels through diet. Because most patients with diabetes are treated with pharmaceutical therapy, the active participation of pharmacists; for example, by instructing the primary caregivers to correctly and consistently administer anti-diabetic drugs, is one way to increase medication compliance and reduce the risk of diabetic complications. In Korea, dementia patients are eligible as beneficiaries of the National Long-Term Care Insurance (NLTCI) and can receive care services from certified care workers. If the NLTCI would determine that supportive diabetes management for dementia patients is the duty of certified care workers, more systematic diabetes management of vulnerable elderly would be possible nationwide.

The authors declare that they have no conflict of interest.

We thank Health Insurance Review and Assessment Service for allowing us to access the 2018 Health Insurance Review and Assessment Service-Adult Patient Sample (2018 HIRA-APS, Serial number: HIRA-APS-2018-0043).