Disparities in Antihypertensive Medication Adherence in Persons With Disabilities and Without Disabilities: Results of a Korean Population-Based Study

Data Source 

Analyses were conducted using the NHI database (2004) and the National Disability Database. The latter comprises approximately 77.7% of the population with disabilities registered in the National Social Security System.10

In South Korea, the NHI provides mandatory universal health insurance, offering comprehensive medical care coverage to all South Korean residents. Patients should visit a doctor to obtain a prescription and then visit a pharmacy to fill that prescription, according to the prescription-dispensation separation policy instituted in 2000. The NHI pharmacy claims are generated when a patient uses insurance coverage to fill a prescription. The claims are then entered into the NHI's claims database. We could not include the information of Medicaid beneficiaries, who account for about 3.1% of all Korean residents, because of the lack of information in the NHI claims database. The Medicaid program is a public assistance scheme to secure the minimum livelihood for low-income households and to provide them with medical services. We used pharmacy claims data and enrollment records from January 1, 2004, to December 31, 2004. The data comprised patients' disease codes, the dates that prescriptions were generated and filled, and the days' worth of pills supplied. Enrollment records included patients' insurance identification numbers and information regarding sex, age, contributions a month, and residential region. These were linked to the pharmacy claims data.

Persons with disabilities were identified using the National Disability Registry (2003), which is a voluntary self-report system for persons with disabilities and is linked with the pharmacy claims data. This database is assumed to cover 77.7% of the total population with disabilities.10 It contains social security numbers and information regarding disability characteristics such as disability type and severity. Disabilities are classified into 15 groups, and diagnosis is made by the respective medical specialist. The severity of the disability is graded from 1 to 6, indicating very severe to very mild, on the basis of functional losses and clinical impairment, as determined by the medical specialist. Diagnostic criteria are specified in the KDA (table 1). The personal identification number used for data linkage was deleted before analysis. In addition, this study was a secondary data analysis. For these reasons, this study was inapplicable to the review from the scientific review panel.

Table 1. Disablement Characteristics According to the KDA

	Category	Disabling Condition	Definition
	External physical disability	Physical impairment	Amputation, motor disturbance, joint disability, deformity of limbs, spinal cord injury
		Brain impairment	Brain disability caused by stroke, brain damage, brain palsy
		Visual impairment	Visual power loss, visual field defect
		Auditory impairment	Hearing disability, disability of the sense of equilibrium
		Linguistic impairment	Mogilalia, dysphonia
		Facial deformity	Facial deformity caused by head injury, a burn, cancer surgery
	Internal organ disability	Renal function impairment	Renal impairment with hemodialysis, kidney transplantation
		Heart function impairment	Heart impairment with daily life limitation
		Liver function impairment	Liver impairment with daily life limitation
		Respiratory function impairment	Respiratory impairment with daily life limitation
		Intestinal and urinary tract function impairment	Intestinal and urinary tract impairment with daily life limitation
		Epilepsy	Epilepsy with daily life limitation
	Mental disability	Mental retardation	Intelligence quotient <70
		Developmental disability	Autism, developmental impairment
		Mental impairment	Mental diseases with daily life limitation

NOTE. The level of severity for each condition ranges from 1 (severe) to 6 (mild).

Study Population 

The source population consisted of patients over the age of 20 years who visited a doctor more than once for hypertension (International Classification of Diseases and Related Health Problems–10th Revision), as reported in claims data from January to December 31, 2004. Accordingly, the target population consisted of 2,716,682 subjects. This corresponded to 28.7% of the 9,475,043 NHI beneficiaries who were over 20 years of age.

In the target population, patients who filled only 1 prescription for an antihypertensive were excluded because the CMA can be calculated only when a prescription has been filled more than twice. Patients who had been hospitalized during the study period were also excluded because adherence to a medication regimen during hospitalization is enforced by the medical staff and requires no effort on the part of the patient. On the basis of these criteria, the study population comprised 2,453,734 patients, which was 90.3% of the source population. In this group, there were 85,098 persons with disabilities and 2,368,636 persons without disabilities.

Outcome Measures 

Adherence to a medication regimen was measured using the CMA scale, which is based on the pharmacy claims data.11 The CMA is defined as the percentage of days for which the patient has pills available. This proportion is calculated by dividing the total number of days of medication dispensed (excluding the final prescription) by the total number of days between the first and last prescriptions. For example, patients who filled 3 hypertension prescriptions for 30 pills each, with the last prescription filled 90 days after the first one, would have a CMA of 67% (ie, 60/90×100) because they had 60 pills available during the 90 days between the first and last prescriptions. Pills from the last prescriptions are not included because the number of days taken to consume these pills is unknown. Therefore, theoretically, the denominator or the total number of days between the first and last prescriptions can range from 1 day to 12 months. For the numerator, the total number of days of medication dispensed (excluding the final prescription) can range from 1 day to over 12 months because patients get the prescription before finishing the previous prescription (prescription duplication). Therefore, CMA can range from 0% to over 100% because of prescription duplication. A CMA greater than or equal to 80% was considered appropriate medication adherence because it is a commonly used threshold to define antihypertensive medication adherence.12, 13 Haynes et al14 reported that hypertension of those who were adherent over 80% of antihypertensive medication was better controlled than those who were adherent under 80%.

Adjustment Factors 

We reviewed studies that evaluated factors that influence medication adherence, and the appropriate adjustment factors were selected.11, 12, 13, 15 These factors consisted of patient demographic characteristics such as sex and age; clinical characteristics such as medication duration and comorbid conditions; socioeconomic characteristics such as type of medical insurance (self-employed vs employee) and insurance contributions a month as a proxy for household income; and community factors such as type of residential area (metropolitan city, city, rural area, mostly rural area). In previous studies, sex was reported to be related with medication adherence in some studies, while in other studies, it was not.16, 17, 18 As for age, several studies reported the trend of increasing medication adherence with increasing age.17, 18, 19, 20 Patients' socioeconomic level,21 patients' residential area,22, 23 duration of medication,21 and patients' comorbidity21 were also reported to be related with medication adherence previously.

Comorbid conditions were classified into 2 disease groups. Class 1 comprised hypertension-related diseases—for example, diabetes mellitus, obesity, and target organ damage caused by hypertension, as defined by the Joint National Committee.1 Class 2 comprised chronic diseases, which were assumed to increase both the number of pharmacy visits and overall medication adherence. Diseases in this group included migraine and other headache syndromes, chronic lower respiratory diseases, and inflammatory polyarthropathies.

Statistical Analysis 

Because the distribution of CMA was not normative, the median and interquartile range of CMAs and the medication adherence rates of persons with and without disabilities were calculated according to the type and severity of the disability. The type of disability was coded as total disability if study subjects were registered in the National Disability Registry and coded as no disability if study subjects were not. In these processes, various types of disability and disease severity were also coded using information in the National Disability Registry. A severe disability was scored as 1 to 3 and a mild disability as 4 to 6. To determine whether various types of disability and severity of disability have influence on the medication adherence, we used the multiple logistic regression model, adjusting for confounding factors and setting the outcome (ie, appropriate medication adherence defined as CMA ≥80%) as a binary variable. Various types of disability and severity of disability were entered in the multiple logistic regression model 1 by 1, adjusting the same confounding factors. Estimates were adjusted for demographic characteristics (sex, age), insurance contributions a month as a proxy for household income, residential area, and clinical characteristics (medication duration, comorbid conditions). All analyses were performed using SAS 81.a Statistical significance was defined as 2-tailed P value less than or equal to .05.

Characteristics of Persons With Disabilities and Without Disabilities 

We summarized the general characteristics of the persons with disabilities and those without disabilities (table 2). Of the 2,453,734 persons in the study, those with disabilities accounted for 3.47% (n=85,098). Male sex occurred more frequently in the population with disabilities than in those without disabilities (61.1% vs 41.2%). In addition, the proportion of self-employed and the number of people with comorbid conditions were higher in the population with disabilities (66.3% vs 43.6% and 65.2% vs 56.4%, respectively). The household income of people with disabilities was lower than that of those without disabilities.

Table 2. Characteristics of the Persons With and Without Disabilities

	Variable	People Without Disabilities		People With Disabilities
		No.	%	No.	%
	Total	2,368,636	100.00	85,098	100.00
	Sex
	Men	975,452	41.18	51,985	61.09
	Women	1,393,184	58.82	33,113	38.91
	Age (y
	20–29	7554	0.32	137	0.16
	30–39	62,370	2.63	1310	1.54
	40–49	327,419	13.82	8971	10.54
	50–59	628,774	26.55	22,312	26.22
	60–69	734,103	30.99	31,526	37.05
	≥70	608,416	25.69	20,842	24.49
	Type of medical insurance
	Self-employed	1,032,762	43.60	56,453	66.34
	Employee	1,335,874	56.40	28,645	33.66
	Comorbidity
	None of class 1 or 2 diseases	1,047,617	44.23	30,165	35.45
	Class 1⁎	911,924	38.50	41,407	48.66
	Class 2†	272,785	11.52	8277	9.73
	Both class 1 and 2 diseases	136,310	5.75	5249	6.17
	Medication duration (mo)
	<6	162,808	6.87	4829	5.67
	6–11	266,516	11.25	8479	9.96
	12–17	275,845	11.65	9392	11.04
	≥18	1,663,467	70.23	62,398	73.32
	Contribution a month (Korean won)
	<30,000	755,849	31.91	32,986	38.76
	30,000–70,000	975,384	41.18	33,890	39.82
	>70,000	637,403	26.91	18,222	21.41
	Residential area
	Metropolitan area	1,119,310	47.26	36,518	42.91
	City	990,661	41.82	37,505	44.07
	Rural	102,529	4.33	4496	5.28
	Mostly rural	156,130	6.59	6579	7.73

⁎ Class 1: diabetes mellitus and target organ damage caused by hypertension. † Class 2: chronic disease assumed to increase ambulatory service use.

CMAs According to Disability Condition and Severity 

The median and interquartile range of CMAs were calculated by classifying the population with disabilities according to disability type and severity (table 3). Except for the group with some types of mental and internal organ disability, most people with disability had lower CMAs than those without. Furthermore, people with more severe disability had lower CMAs than those with mild disability for disabilities such as brain impairment, auditory impairment, mental retardation, and epilepsy.

Table 3. CMA by Disabling Condition and Severity

	Variable	Total				Mild Disability				Severe Disability
		No.	Median	IQR	≥80% CMA	No.	Median	IQR	≥80% CMA	No.	Median	IQR	≥80% CMA
	Persons without disabilities	2,368,636	85.71	38.34	57.53			−				−
	Persons with disabilities	85,098	83.56	42.13	54.54	49,943	83.33	41.35	54.56	35,155	83.78	43.28	54.51
	External disability
	Limb impairment	50,070	83.45	41.52	54.63	33,155	83.28	41.34	54.41	16,915	83.92	41.87	55.04
	Brain impairment	12,014	85.94	40.57	57.08	2764	87.51	38.26	59.23	9250	85.47	41.65	56.44
	Visual impairment	9846	83.28	42.01	54.34	7529	83.07	41.68	54.24	2317	83.57	43.06	54.64
	Auditory impairment	8564	81.68	44.43	52.01	5311	82.95	43.26	53.42	3253	79.72	46.26	49.71
	Linguistic impairment	1833	84.98	40.49	56.79	856	84.55	40.11	56.66	977	85.71	40.78	56.91
	Facial deformity	20	85.14	34.21	60.00	10	85.14	18.73	50.00	10	85.14	18.73	70.00
	Mental disability
	Mental retardation	786	84.94	44.69	53.82	58	87.93	35.19	56.90	728	84.85	45.18	53.57
	Developmental disability	10	82.33	27.02	50.00	2	78.48	0.92	0.00	8	85.96	39.95	62.50
	Mental disease	171	87.46	43.44	60.23	3	97.92	11.48	100.00	168	86.75	43.77	59.52
	Internal organ disability
	Renal impairment	557	59.65	73.22	35.73	68	69.83	64.40	41.18	489	58.61	74.89	34.97
	Heart impairment	782	82.87	49.52	52.56	3	97.34	215.27	66.67	779	82.82	49.52	52.50
	Respiratory impairment	204	76.16	46.31	45.10	0	—	—	—	204	76.16	46.31	45.10
	Liver impairment	44	51.29	62.24	31.82	6	38.08	54.56	33.33	38	53.66	65.82	31.58
	Intestinal and urinary tract impairment	178	84.66	41.65	53.37	166	84.30	41.02	53.01	12	88.71	51.32	58.33
	Epilepsy	19	56.36	71.27	26.32	12	58.18	71.22	33.33	7	56.36	44.60	14.29

Abbreviation: IQR, interquartile range.

Disparity in Antihypertensive Medication Adherence According to Disability Condition and Severity 

The multiple logistic regression adjusting other demographic and clinical variables (table 4) indicated that disability was associated with a lower CMA (AOR=.87; 95% CI, .86–.88). In addition, people with severe disabilities were more likely to have a lower CMA than those with mild disability. Severe limb (AOR=.87; 95% CI, .85–.90), brain (AOR=.90; 95% CI, .86–.94), visual (AOR=.88; 95% CI, .81–.95), auditory (AOR=.76; 95% CI, .70–.81), and internal organ (AOR=.58; 95% CI, .53–.65) impairments were associated with a lower appropriate medication adherence. Higher disability severity was associated with lower CMA. For subjects disabled with a facial deformity or a linguistic, intellectual, or chronic mental impairment, there were no disparities in medication adherence compared with the population without disabilities.

Table 4. AOR and 95% CIs for Logistic Regression of CMA by Disabling Condition and Severity

	Disabling Condition	Severity	OR	95% CI	AOR⁎	95% CI⁎
	No disability		1		1
	Disability (total)	Severe disability	0.88	0.87–0.90	0.86	0.84–0.88
		Mild disability	0.89	0.87–0.90	0.88	0.87–0.90
		Total	0.89	0.87–0.90	0.87	0.86–0.88
	External physical disability
	Limb impairment	Severe disability	0.90	0.88–0.93	0.87	0.85–0.90
		Mild disability	0.88	0.86–0.90	0.88	0.86–0.90
		Total	0.89	0.87–0.91	0.88	0.86–0.89
	Brain impairment	Severe disability	0.96	0.92–0.99	0.90	0.86–0.94
		Mild disability	1.07	0.99–1.16	0.98	0.91–1.05
		Total	0.98	0.95–1.02	0.92	0.88–0.95
	Visual impairment	Severe disability	0.89	0.82–0.96	0.88	0.81–0.95
		Mild disability	0.87	0.84–0.92	0.87	0.83–0.91
		Total	0.88	0.84–0.91	0.87	0.84–0.91
	Auditory impairment	Severe disability	0.73	0.68–0.78	0.76	0.70–0.81
		Mild disability	0.85	0.80–0.90	0.86	0.81–0.91
		Total	0.80	0.77–0.84	0.82	0.78–0.85
	Linguistic impairment	Severe disability	0.98	0.86–1.11	0.95	0.82–1.06
		Mild disability	0.97	0.84–1.11	0.93	0.81–1.07
		Total	0.97	0.89–1.06	0.93	0.85–1.03
	Facial deformity	Severe disability	1.67	0.44–6.39	1.64	0.42–6.43
		Mild disability	0.74	0.21–2.54	0.69	0.20–2.39
		Total	1.11	0.45–2.71	1.04	0.42–2.57
	Mental disability
	Intellectual disability	Severe disability	0.85	0.74–0.99	0.92	0.79–1.07
		Mild disability	0.90	0.54–1.50	0.86	0.52–1.45
		Total	0.89	0.55–1.37	0.91	0.72–1.05
	Mental disease	Total	1.22	0.85–1.59	1.15	0.84–1.57
	Internal organ disability†	Severe disability	0.61	0.55–0.68	0.58	0.53–0.65
		Mild disability	0.70	0.55–0.89	0.71	0.55–0.91
		Total	0.62	0.57–0.68	0.60	0.55–0.66

⁎ Values were adjusted by sex, age, type of medical insurance, contribution a month, comorbidity, residential area, and duration of medication. † Renal function impairment, heart function impairment, liver function impairment, respiratory function impairment, intestinal and urinary tract function impairment, epilepsy.

Study Limitations 

Several limitations of our study should be mentioned. First, the Medicaid beneficiaries, who account for about 3.1% of all Korean residents, were excluded because of the lack of information in the NHI claims database.26 Nonetheless, except for those with mental disabilities, Medicaid beneficiaries account for below 10% of the persons with disabilities. Even in persons with mental disabilities, Medicaid beneficiaries account for below 20%. Therefore, the exclusion of this segment of the population with disabilities may have reduced the study's representativeness, but these problems could not affect the internal validity of the study results. Another limitation of this study is that the coverage rate of the National Disability Registry is 77.7%, reflective of the fact that it is a voluntary, self-reporting system for people with disabilities. However, it can be assumed that most persons with mobility or functional impairments according to the types of disability were included in the registry, because the government provides various benefits such as income aid and financial aid for medical services. Although the problem of representativeness of the study subjects exists, this study covers persons with every type of disability and severity, so it could not have affected the internal validity of the study results. Also, we excluded those who had been hospitalized during the study period. This group of subjects accounted for a small fraction compared with the target population, and because the concept of medication adherence is not traditionally considered in hospitalization settings, it could not have affected the study results as a whole. Second, the assessment of adherence was based on pharmacy refill data, which can overestimate adherence because the filling of a prescription does not confirm the actual ingestion of the prescribed medication. Refill adherence therefore identifies the upper boundary of medication consumption. However, because patients with an insufficient number of pills prescribed for a given period cannot fully adhere to their medication regimens, we believe that our method is well suited to identifying suboptimal adherence over time.39, 40 A disability may lower the probability of appropriate ingestion of medication, which would widen the disparity in adherence between persons with disabilities and without disabilities. Third, this study could not include other potential confounding factors reported in the previous studies. Side effects of treatment, patients' knowledge and attitude, and different treatment modalities have been reported to affect medication adherence.41 Last, because of the relatively small sample size in specific disability types (especially mental diseases), there is a possibility that we could not identify true disparities between the 2 groups. In the future, larger and more focused studies dealing with specific disability types should be conducted.

The strength of our study is that, in contrast with other adherence-monitoring methods such as self-reporting, pill counts, and electronic monitoring, patients were not aware that adherence was being measured and therefore did not alter their usual behavior during the observation period.31 Thus, disparities in antihypertensive medication adherence between people with disabilities and those without could be evaluated using an objective methodology.