Continuity of Care and Risks of Hospitalization and Mortality in COPD: A Nationwide Cohort Study

Abstract

Purpose

Chronic obstructive pulmonary disease (COPD) is characterized by unpredictable patterns of exacerbations, making continuity of care (CoC) a critical component in disease management. CoC influences COPD exacerbation-related hospitalization and mortality. Therefore, we assessed the impact of CoC levels on subsequent hospital admissions and all-cause mortality among individuals with COPD.

Patients and Methods

This retrospective nationwide cohort study utilized National Health Insurance Service-Senior cohort data (2002–2019) and included 29,316 patients newly diagnosed with COPD. The primary exposure was longitudinal CoC level categorized as low (<0.7) or high (≥0.7). Alternative continuity indices (usual care provider, sequential continuity, and modified continuity indices) were also evaluated. Outcomes were COPD exacerbation-related hospitalization and all-cause mortality within 1 year of diagnosis. Cox proportional hazards models estimated hazard ratios (HRs), Kaplan–Meier curves, and cumulative incidence rates were used to assess outcomes, and the Log rank test was used for between-group comparisons.

Results

Low CoC level were associated with an increased risk of 3-year COPD exacerbation-related hospitalization (HR 1.63, 95% CI 1.45–1.83) and all-cause mortality (HR 1.25, 95% CI 1.11–1.40) compared with high CoC level. Intermediate (0.4–0.7) and low (<0.4) CoC level showed progressively increased hospitalizations (HR 1.61, 95% CI 1.40–1.86 vs HR 1.65, 95% CI 1.38–1.97), and mortality (HR 1.24, 95% CI 1.07–1.43 vs HR 1.26, 95% CI 1.05–1.51), respectively. These findings remained consistent across alternative continuity indices. Low CoC level were associated with an increased risk of emergency (HR 1.48, 95% CI 1.19–1.83) and general hospital admissions (HR 1.70, 95% CI 1.47–1.95).

Conclusion

Lower CoC scores level consistently associated with higher risks of COPD hospitalization and all-cause mortality across multiple continuity indices. Subgroup and sensitivity analyses indicated that fragmented outpatient care increased adverse outcomes. Strengthening longitudinal patient–provider relationships may reduce preventable hospitalizations and premature deaths in patients with COPD.

Introduction

Chronic obstructive pulmonary disease (COPD) ranks among the leading causes of death worldwide, accounting for 74.4 million disability-adjusted life years globally in 2019,¹ with a prevalence of 13.4% among adults aged ≥40 years in South Korea.² Acute exacerbations increase rehospitalization rates (12.7%)³ and healthcare costs (1245 million USD).⁴ This issue is particularly evident among older adults aged ≥65 years, where comorbidities (eg, heart failure, diabetes) and frailty further elevate hospitalization and mortality risks Beyond airflow limitation, COPD is also associated with cognitive impairment,⁵ malnutrition,⁶ and social isolation, underscoring the urgent need for effective management strategies.⁷

Continuity of care (CoC), defined as an ongoing therapeutic relationship with the same provider, is critical in COPD management owing to the disease’s chronic, unpredictable exacerbation pattern. In large-scale population-based research, claims-based indices such as the Bice-Boxerman index are widely used to evaluate this concept. However, these indices primarily capture the concentration and dispersion of outpatient visits. Therefore, while they robustly reflect management and informational aspects of continuity, they may not fully represent the qualitative nuances of interpersonal relational continuity.

In the South Korean healthcare system, universal coverage is provided under a fee-for-service model with no strict primary care gatekeeping mechanism. Patients have unrestricted access to any healthcare provider, including specialists and tertiary hospitals. While this ensures high accessibility, the absence of a designated primary care coordinator often leads to doctor-shopping behavior, structurally driving the fragmentation of care.

Within this context, COPD serves as an optimal test case to evaluate the impact of CoC. COPD is a quintessential ambulatory care-sensitive condition that requires long-term, coordinated management, including regular lung function monitoring, continuous patient education on inhaler techniques, and lifestyle modifications. Because effective management of COPD relies heavily on a consistent therapeutic relationship to prevent acute exacerbations, the clinical outcomes (eg, hospitalizations and mortality) of these patients are highly sensitive to fragmented care.

To categorize patients with high CoC, we adopted a cut-off of 0.7, a criterion utilized in several prior studies to denote highly concentrated outpatient care.⁸ A high CoC level (CoC index ≥0.7) enhances medication adherence (inhaled corticosteroids and bronchodilators),⁹ enables early symptom detection (cough and dyspnea), and facilitates personalized interventions, (smoking cessation, nutritional counseling, and vaccination) to reduce exacerbation frequency.^10–12 In frail older patients with COPD, where polypharmacy and social isolation impair self-management, a high CoC level enables multidisciplinary coordination to mitigate compounded risks.^13,14 Each hospitalization for a COPD exacerbation is a strong predictor of subsequent rehospitalization^3,15 and accounts for a large proportion of COPD-related healthcare costs, which are primarily due to inpatient and emergency care.^4,16 Therefore, strengthening CoC to prevent exacerbations offers substantial potential to improve long-term survival and reduce COPD-related healthcare spending by mitigating expensive hospital admissions.^15,17

Despite this body of evidence, most previous work has been conducted in Western health systems with strong primary-care gatekeeping, and relatively few studies have evaluated CoC using person-level indices in Asian or other non-gatekeeping contexts.^18,19 South Korea’s rapidly aging population, projected to become a super-aged society by 2025, and the escalating burden of COPD provide a compelling rationale to address this gap. To capture nationwide patterns, this study utilized the National Health Insurance Service (NHIS)-Senior cohort. We acknowledge that evaluating COPD outcomes using administrative claims data inherently introduces challenges related to clinical heterogeneity and unmeasured confounding; critical details such as spirometry results (eg, FEV1) and daily symptom burden cannot be fully captured and may influence both care patterns and outcomes. Despite these limitations, this nationwide resource is essential for substantiating CoC-centered primary care enhancements and informing national reimbursement reforms. Therefore, this study aimed to investigate the impact of CoC levels on subsequent hospital admissions and all-cause mortality among older patients with COPD, thereby clarifying how outpatient continuity influences clinical outcomes in a free-access health system.

Materials and Methods

Data

The data analyzed in this study were obtained from the NHIS-Senior Cohort from 2002 to 2019. The NHIS database is used to establish health policies and conduct biomedical research, providing researchers with comprehensive data on the claims collected.²⁰ The cohort was constructed to represent older individuals living in South Korea. Specifically, 10% of 5.5 million subscribers aged ≥60 years were randomly selected and enrolled through simple random sampling.²⁰ All individuals included in the NHIS-Senior cohort were followed up until December 31, 2019, unless they were disqualified from NHIS coverage owing to certain factors, such as death or emigration.²¹ The NHIS-Senior Cohort also includes information on death status and date of death, enabling ascertainment of all-cause mortality during follow-up. The NHIS is the sole insurer under the nationwide health insurance single-payer system in South Korea. Most citizens are NHIS-affiliated and are categorized as insured, self-employed, and healthcare benefit recipients. Key variables in the NHIS-Senior database include hospitalization and outpatient billing data, such as procedure codes, prescriptions, and diagnoses,^21,22 as well as socioeconomic status information and clinically determined International Classification of Diseases 10^th revision (ICD-10) codes.

The data supporting the findings of this study are available from the National Health Insurance Sharing Service. However, access to these data is restricted and requires approval. Data are available for use after approval by the NHIS (https://nhiss.nhis.or.kr/bd/ab/bdaba000eng.do).

Participants

The study population was selected according to the following criteria. To select participants newly diagnosed with COPD, a 2-year washout period was applied. Baseline COPD was defined using ICD-10 code J44 recorded as either a primary or secondary diagnosis, based on at least one hospital admission or three or more outpatient visits. To calculate the CoC and mitigate immortal-time bias, we applied a 1-year landmark period. Specifically, the CoC measurement window was defined as the 1-year (365 days) period immediately following the initial COPD diagnosis. To ensure the mathematical stability of the CoC index and prevent extreme distortion caused by an excessively small denominator (eg, only one or two visits), participants with fewer than three outpatient visits during the 1-year measurement window were excluded.²³ Additionally, those who died or experienced the primary outcome before the end of this window were also excluded to fulfill the landmark analysis criteria. Following this 1-year landmark period, the eligible participants were followed up for up to 3 years to evaluate the occurrence of COPD exacerbation-related hospitalizations and all-cause mortality.

Variables

The variable of interest was the longitudinal CoC index, calculated using the Bice–Boxerman formula. This index quantifies the extent to which a patient visits the same provider across outpatient visits following COPD diagnosis and prior to outcome occurrence, ranging from 0 (visits dispersed across multiple providers) to 1 (all visits with a single provider). In this formula, denotes the total number of distinct providers (defined as hospitals or clinics with unique identifiers), represents the number of visits to the -th provider, and indicates the overall number of outpatient visits across all providers during the observation period. This measure reflects the concentration of care, with higher values indicating greater continuity when visits are concentrated among fewer providers.

Bice-Boxerman CoC index =

Based on previous studies, CoC levels were categorized as low (<0.7) or high (≥0.7).⁸ A CoC level of 0.7 or higher reflects a relatively concentrated pattern of outpatient care, indicating that most visits are provided by the same provider rather than dispersed across multiple providers. To ensure meaningful longitudinal continuity measurement, patients with less than three outpatient visits during the follow-up period were excluded.

For sensitivity analysis, CoC level were further categorized into three subgroups: low (<0.4), intermediate (0.4 and 0.7), and high (≥0.7). Additionally, three alternative continuity indices were examined: the usual provider of care index (UPC), sequential continuity index (SECON), and modified continuity index (MMCI). The UPC index was calculated as the proportion of total outpatient visits made to the single provider (hospital or clinic) with the highest number of visits during the observation period, yielding values from 0 (visits dispersed across multiple providers) to 1 (all visits with one provider). SECON was defined as the proportion of consecutive visit pairs in which the same provider was seen, thereby capturing the stability of provider sequence over time, with values ranging from 0 to 1. The MMCI was calculated using the total number of outpatient visits and the number of distinct providers seen, expressed as (N−M)/(N−1) where N is the number of visits and M is the number of different providers, such that higher level indicate greater concentration of visits among fewer providers.

UPC =

SECON =

MMCI =

Because the CoC, UPC, SECON, and MMCI indices are standardized measures ranging from 0 to 1, the same cut-off values (<0.4, 0.4–0.7, and ≥0.7) were applied across all indices to facilitate comparison and provide a consistent framework for interpreting continuity levels.

The primary outcomes were (1) COPD exacerbation-related hospitalizations and (2) all-cause mortality within 3 years following the 1-year landmark period. Hospitalizations due to COPD exacerbations were defined as inpatient admissions with ICD-10 codes J44, and further stratified based on the mode of hospital entry (emergency department vs general admission).

The covariates included patient characteristics, namely sex (male, female), age (<65, 65–69, 70–74, 75–79, ≥80 years), region (metropolitan, city, rural), income level (low, middle, high), type of medical insurance (workplace insured, regionally insured, medical aid), disability status (yes, no), Charlson Comorbidity Index (CCI) score (0–1, 2–3, ≥4), and COPD severity (mild, moderate, severe), the number of outpatient visits during the previous year (categorized into quartiles: Q1 to Q4) as well as hospital characteristics, namely hospital location (metropolitan, city, rural) and hospital classification (general hospital, hospital, clinic, other). We operationally defined baseline COPD severity into three categories (Mild, Moderate, and Severe) based on the intensity of COPD-specific medication regimens prescribed for ≥5 cumulative days during the baseline period. The Mild category included patients without prescriptions for long-acting muscarinic antagonists (LAMA), long-acting β2-agonists (LABA), inhaled corticosteroids (ICS), or fixed-dose combinations (COMB). The Moderate category was defined as maintenance monotherapy with either LAMA or LABA. The Severe category represented step-up therapy, including dual bronchodilators (LAMA plus LABA), COMB, ICS-containing regimens, or the addition of oral medications (leukotriene receptor antagonists or theophylline) to monotherapy.

Statistical Analysis

A chi-square test was performed to assess the frequency and percentage of each categorical variable by patient at baseline and the distribution of hospital admission and all-cause mortality according to each variable. Cox proportional hazards regression models were employed as the primary analytical approach to evaluate the association between the CoC indices and the risk of primary outcomes (COPD exacerbation-related hospitalizations and all-cause mortality). The proportional hazards assumption was evaluated using Schoenfeld residuals. The assumption was met for our primary exposure variable, the CoC index (p > 0.05). Furthermore, because death precludes the occurrence of hospitalization, a competing risk analysis was conducted using the Fine and Gray subdistribution hazard model. In this model, all-cause mortality was treated as a competing event for COPD exacerbation-related hospitalization. Kaplan–Meier survival curves were plotted to compare the cumulative incidence of events across CoC categories, and the Log rank test was used to evaluate between-group differences. For sensitivity analyses, CoC was recategorized into three levels (low: <0.4, intermediate: 0.4–0.7, high: ≥0.7), and alternative continuity indices (UPC, SECON, and MMCI) were also evaluated.

Ethical Consideration

This study was reviewed and approved by the International Deliberative Committee on the Health Care System of Yonsei University (Registration No.: 4–2025-1526). The committee waived the requirement for informed consent because the database used was based on routine administrative and billing data. All procedures contributing to this work comply with relevant national and institutional committees on human experiments and the ethical standards of the Declaration of Helsinki 1975, as amended in 2008.

Results

Table 1 summarizes the general characteristics of the study population based on the univariate analysis. A total of 29,316 individuals were included, of whom 6.1% experienced hospital admission and 7.3% died during follow-up. During the 3-year follow-up, the Kaplan–Meier survival curves for all-cause mortality showed slightly higher survival among patients with a high CoC level (≥0.7) compared with those with a low CoC level (<0.7) (Figure 1). The cumulative incidence rate during the study period differed significantly depending on the CoC level (log-rank p < 0.0001). The 3-year cumulative incidence curves for all-cause hospitalization demonstrated a consistently lower risk of admission in the high CoC level group than in the low CoC level group (Figure 2). The gap between the curves widened over time, indicating fewer hospitalizations among patients with a CoC level ≥0.7.

Table 1.

General Characteristics of the Study Population

Variables	Hospital Admission							All-Cause Mortality
	Total		Yes		No		P-value	Yes		No		P-value
Total (N=29,316)	N	%	N	%	N	%		N	%	N	%
	29,316	100	1773	6.0	27,543	94.0		2128	7.3	27,188	92.7
CoC index							<0.0001					<0.0001
≥0.7	25,947	88.5	1411	79.6	24,536	89.1		1790	84.1	24,157	88.9
<0.7	3369	11.5	362	20.4	3007	10.9		338	15.9	3031	11.1
Sex							<0.0001					<0.0001
Male	17,561	59.9	1363	76.9	16,198	58.8		1569	73.7	15,992	58.8
Female	11,755	40.1	410	23.1	11,345	41.2		559	26.3	11,196	41.2
Age							<0.0001					<0.0001
<65	8571	29.2	364	20.5	8207	29.8		174	8.2	8397	30.9
65–69	6204	21.2	335	18.9	5869	21.3		313	14.7	5891	21.7
70–74	6603	22.5	441	24.9	6162	22.4		498	23.4	6105	22.5
75–79	5160	17.6	401	22.6	4759	17.3		612	28.8	4548	16.7
≥80	2778	9.5	232	13.1	2546	9.2		531	25.0	2247	8.3
Region							<0.0001					0.0063
Metropolitan	9695	33.1	518	29.2	9177	33.3		683	32.1	9012	33.1
City	6752	23.0	378	21.3	6374	23.1		445	20.9	6307	23.2
Rural	12,869	43.9	877	49.5	11,992	43.5		1000	47.0	11,869	43.7
Income							0.5016					0.0001
Low	8846	30.2	557	31.4	8289	30.1		716	33.6	8130	29.9
Middle	8665	29.6	515	29.0	8150	29.6		557	26.2	8108	29.8
High	11,805	40.3	701	39.5	11,104	40.3		855	40.2	10,950	40.3
Type of medical insurance							<0.0001					<0.0001
Workplace insured	8926	30.4	486	27.4	8440	30.6		566	26.6	8360	30.7
Regionally insured	17,651	60.2	1057	59.6	16,594	60.2		1253	58.9	16,398	60.3
Medical aid	2739	9.3	230	13.0	2509	9.1		309	14.5	2430	8.9
Disorder							<0.0001					<0.0001
No	24,237	82.7	1402	79.1	22,835	82.9		1587	74.6	22,650	83.3
Yes	5079	17.3	371	20.9	4708	17.1		541	25.4	4538	16.7
CCI							0.0013					<0.0001
0–1	11,390	38.9	627	35.4	10,763	39.1		488	22.9	10,902	40.1
2–3	11,133	38.0	682	38.5	10,451	37.9		734	34.5	10,399	38.2
≥4	6793	23.2	464	26.2	6329	23.0		906	42.6	5887	21.7
COPD severity							<0.0001					0.0028
Mild	22,916	78.2	1093	61.6	21,823	79.2		1618	76.0	21,298	78.3
Moderate	3104	10.6	337	19.0	2767	10.0		272	12.8	2832	10.4
Severe	3296	11.2	343	19.3	2953	10.7		238	11.2	3058	11.2
Number of outpatient visits during the previous year							0.012					<0.0001
Q1 (Low)	6673	22.8	359	20.2	6314	22.9		340	16.0	6333	23.3
Q2	7224	24.6	417	23.5	6807	24.7		437	20.5	6787	25.0
Q3	7680	26.2	497	28.0	7183	26.1		584	27.4	7096	26.1
Q4 (High)	7739	26.4	500	28.2	7239	26.3		767	36.0	6972	25.6
Hospital location							0.0064					0.0588
Metropolitan	10,265	35.0	567	32.0	9698	35.2		720	33.8	9545	35.1
City	7347	25.1	440	24.8	6907	25.1		507	23.8	6840	25.2
Rural	11,704	39.9	766	43.2	10,938	39.7		901	42.3	10,803	39.7
Hospital classification							<0.0001					<0.0001
General hospital	11,009	37.6	1026	57.9	9983	36.2		940	44.2	10,069	37.0
Hospital	1931	6.6	193	10.9	1738	6.3		181	8.5	1750	6.4
Clinic	15,899	54.2	531	29.9	15,368	55.8		980	46.1	14,919	54.9
Others	477	1.6	23	1.3	454	1.6		27	1.3	450	1.7

Notes: Data are presented as number for categorical variables. P-values were calculated using Pearson’s chi-square test.

Abbreviations: CCI, Charlson Comorbidity Index; CoC, Continuity of Care; COPD, Chronic Obstructive Pulmonary Disease.

Figure 1.

Kaplan–Meier survival curves for 3-year all-cause mortality by continuity of care level (CoC ≥0.7 vs CoC <0.7).

Figure 2.

Cumulative incidence of 3-year hospitalization by continuity of care level (CoC ≥0.7 vs CoC <0.7).

The results from the Cox proportional hazards regression model for hospital admission and mortality risks according to the CoC index and major covariates are presented in Table 2. Compared with patients with a high CoC level, those with a low level had a significantly increased risk of hospital admission (HR 1.63, 95% CI 1.45–1.83) and death (HR 1.25, 95% CI 1.11–1.40). To account for the potential competing risk of death, an additional competing-risk analysis was performed for hospital admission. The results were largely unchanged, with a low CoC level remaining significantly associated with an increased risk of hospital admission (HR 1.64, 95% CI 1.52–1.78), indicating that the observed association was robust to the competing risk of mortality.

Table 2.

Association Between Continuity of Care Level and Hospital Admission and Mortality Using the Cox Proportional Hazard Regression Models

Variables	Hospital Admission	All-Cause Mortality
	HR (95% CI)	HR (95% CI)
CoC index
≥0.7	1.00	1.00
<0.7	1.63 (1.45–1.83)	1.25 (1.11–1.40)
Sex
Male	1.00	1.00
Female	0.54 (0.48–0.60)	0.46 (0.42–0.51)
Age
<65	1.00	1.00
65–69	1.31 (1.13–1.52)	2.46 (2.04–2.96)
70–74	1.67 (1.45–1.93)	3.68 (3.09–4.38)
75–79	2.05 (1.77–2.38)	6.05 (5.10–7.18)
≥80	2.39 (2.02–2.83)	10.88 (9.14–12.96)
Region
Metropolitan	1.00	1.00
City	0.99 (0.76–1.27)	0.89 (0.70–1.14)
Rural	1.05 (0.84–1.32)	0.90 (0.73–1.12)
Income
Low	1.04 (0.91–1.18)	1.16 (1.03–1.30)
Middle	1.13 (1.01–1.27)	1.14 (1.02–1.27)
High	1.00	1.00
Type of medical insurance
Workplace insured	1.00	1.00
Regionally insured	1.06 (0.95–1.18)	1.03 (0.93–1.14)
Medical aid	1.53 (1.27–1.85)	1.35 (1.15–1.59)
Disorder
No	1.00	1.00
Yes	1.11 (0.99–1.25)	1.32 (1.19–1.45)
CCI
0–1	1.00	1.00
2–3	1.00 (0.89–1.12)	1.36 (1.21–1.53)
≥4	1.05 (0.93–1.20)	2.55 (2.26–2.87)
COPD severity
Mild	1.00	1.00
Moderate	1.79 (1.58–2.04)	1.14 (1.00–1.31)
Severe	1.87 (1.64–2.12)	0.99 (0.86–1.14)
Number of outpatient visits during the previous year
Q1 (Low)	1.00	1.00
Q2	1.02 (0.88–1.17)	0.94 (0.82–1.09)
Q3	1.11 (0.96–1.28)	0.97 (0.85–1.12)
Q4 (High)	1.10 (0.95–1.28)	1.04 (0.90–1.19)
Hospital location
Metropolitan	1.00	1.00
City	1.21 (0.95–1.55)	1.14 (0.90–1.45)
Rural	1.37 (1.09–1.72)	1.20 (0.96–1.49)
Hospital classification
General hospital	1.00	1.00
Hospital	1.08 (0.92–1.27)	1.06 (0.90–1.25)
Clinic	0.42 (0.37–0.47)	0.77 (0.70–0.85)
Others	0.53 (0.35–0.81)	0.64 (0.43–0.94)

Note: Hazard ratios were adjusted for all variables listed in the table.

Abbreviations: HR, Hazard Ratio; CI, Confidence Interval; CoC, Continuity of Care; CCI, Charlson Comorbidity Index; COPD, Chronic Obstructive Pulmonary Disease.

Table 3 presents the results of the subgroup analysis stratified by independent variables. Compared with the reference group (CoC index ≥0.7), a CoC level <0.7 was generally associated with an increased risk of hospital admission across most subgroups. Significant interactions were observed according to COPD severity (p for interaction = 0.0272), hospital location (p for interaction = 0.0363), and hospital classification (p for interaction < 0.0001). The association between a low CoC level and hospitalization was observed among patients primarily managed in clinics (HR 2.45, 95% CI 2.00–3.00).

Table 3.

Results of Subgroup Analyses Stratified by Independent Variables

Variables	Hospital Admission			All-Cause Mortality
	CoC Index			CoC Index
	≥0.7	<0.7	p-for Interaction	≥0.7	<0.7	p-for Interaction
	HR	HR (95% CI)		HR	HR (95% CI)
Sex			0.3346			0.9007
Male	1.00	1.58 (1.39–1.80)		1.00	1.25 (1.09–1.42)
Female	1.00	1.75 (1.35–2.28)		1.00	1.25 (0.96–1.64)
Age			0.8225			0.8383
<65	1.00	1.45 (1.09–1.94)		1.00	1.40 (0.92–2.12)
65–69	1.00	1.48 (1.12–1.98)		1.00	1.15 (0.83–1.60)
70–74	1.00	1.62 (1.29–2.03)		1.00	1.19 (0.94–1.52)
75–79	1.00	1.80 (1.43–2.27)		1.00	1.28 (1.03–1.58)
≥80	1.00	1.67 (1.21–2.31)		1.00	1.29 (1.01–1.65)
Region			0.2339			0.0252
Metropolitan	1.00	1.30 (1.02–1.67)		1.00	1.60 (1.30–1.97)
City	1.00	1.89 (1.46–2.45)		1.00	1.18 (0.89–1.56)
Rural	1.00	1.70 (1.45–1.98)		1.00	1.09 (0.92–1.29)
Income			0.1338			0.9038
Low	1.00	1.87 (1.52–2.31)		1.00	1.21 (0.98–1.50)
Middle	1.00	1.34 (1.07–1.68)		1.00	1.28 (1.02–1.61)
High	1.00	1.64 (1.37–1.97)		1.00	1.22 (1.02–1.47)
Type of medical insurance			0.0726			0.9897
Workplace insured	1.00	1.37 (1.08–1.73)		1.00	1.24 (0.99–1.56)
Regionally insured	1.00	1.67 (1.44–1.94)		1.00	1.23 (1.05–1.43)
Medical aid	1.00	1.94 (1.40–2.67)		1.00	1.33 (0.96–1.84)
Disorder			0.7165			0.3432
No	1.00	1.64 (1.44–1.87)		1.00	1.28 (1.12–1.46)
Yes	1.00	1.61 (1.24–2.09)		1.00	1.11 (0.87–1.43)
CCI			0.1037			0.1271
0–1	1.00	1.48 (1.20–1.83)		1.00	1.56 (1.22–1.99)
2–3	1.00	1.86 (1.55–2.22)		1.00	1.18 (0.97–1.45)
≥4	1.00	1.46 (1.15–1.85)		1.00	1.17 (0.98–1.41)
COPD severity			0.0272			0.3804
Mild	1.00	1.71 (1.47–1.99)		1.00	1.24 (1.07–1.43)
Moderate	1.00	1.20 (0.89–1.62)		1.00	1.11 (0.80–1.53)
Severe	1.00	1.81 (1.42–2.30)		1.00	1.44 (1.08–1.93)
Number of outpatient visits during the previous year			0.3927			0.6121
Q1 (Low)	1.00	1.38 (1.02–1.86)		1.00	1.38 (1.00–1.90)
Q2	1.00	1.84 (1.46–2.32)		1.00	1.16 (0.89–1.52)
Q3	1.00	1.55 (1.24–1.94)		1.00	1.42 (1.14–1.76)
Q4 (High)	1.00	1.70 (1.37–2.10)		1.00	1.15 (0.94–1.40)
Hospital location			0.0363			0.3736
Metropolitan	1.00	1.24 (0.99–1.57)		1.00	1.43 (1.17–1.75)
City	1.00	1.75 (1.38–2.22)		1.00	1.24 (0.96–1.50)
Rural	1.00	1.85 (1.57–2.19)		1.00	1.13 (0.95–1.36)
Hospital classification			<0.0001			0.3729
General hospital	1.00	1.30 (1.10–1.53)		1.00	1.27 (1.07–1.52)
Hospital	1.00	1.46 (1.04–2.03)		1.00	0.93 (0.63–1.38)
Clinic	1.00	2.45 (2.00–3.00)		1.00	1.28 (1.07–1.54)
Others	1.00	1.92 (0.65–5.63)		1.00	1.20 (0.43–3.32)

Note: Hazard ratios were adjusted for all other variables listed in the table, except for the variable used for stratification.

Abbreviations: HR, Hazard Ratio; CI, Confidence Interval; CoC, Continuity of Care; CCI, Charlson Comorbidity Index; COPD, Chronic Obstructive Pulmonary Disease.

Regarding mortality, the association between a CoC level <0.7 and the risk of death was less pronounced than that observed for hospital admission. However, a significant interaction was identified according to region of residence (p for interaction = 0.0252), with a stronger association observed among patients residing in metropolitan areas (HR 1.60, 95% CI 1.30–1.97) than among those living in non-metropolitan regions.

Sensitivity analyses using alternative categorizations of the CoC index, as well as the UPC, SECON, and MMCI indices, are summarized in Table 4. When the CoC index was classified into three levels, the risk of hospital admission progressively increased from the high CoC level group (≥0.7, reference) to the intermediate (0.4–0.7; HR 1.61, 95% CI 1.40–1.86) and low CoC level groups (<0.4; HR 1.65, 95% CI 1.38–1.97). A similar gradient was observed for death (HR 1.24, 95% CI 1.07–1.43 and HR 1.26, 95% CI 1.05–1.51, respectively). Using the UPC index yielded consistent findings, with higher risks of hospital admission in the intermediate (HR 1.43, 95% CI 1.23–1.66) and low groups (HR 4.12, 95% CI 2.47–6.85) compared with the high CoC level group (≥0.7), and corresponding HRs for death of 1.27 (95% CI 1.09–1.47) and 2.10 (95% CI 1.13–3.91), respectively. Similar patterns for hospital admission were observed using the SECON index, where intermediate (0.4–0.7) and low (<0.4) CoC levels were associated with increased risks (HR 1.55, 95% CI 1.33–1.80 and HR 1.97, 95% CI 1.46–2.67, respectively) relative to a high SECON level (≥0.7); however, the associations with death were not statistically significant (HR 1.04, 95% CI 0.88–1.23 and HR 1.32, 95% CI 0.95–1.83, respectively). The MMCI also showed a comparable relationship for hospital admission, with HRs of 1.44 (95% CI 1.21–1.70) and 2.06 (95% CI 1.24–3.44) for the intermediate and low CoC level groups, respectively, while the associations with death were also not statistically significant (HR 1.16, 95% CI 0.98–1.38 and HR 1.13, 95% CI 0.62–2.05, respectively).

Table 4.

Sensitivity Analysis of Continuity of Care Using COC Index (0.4 and 0.7), UPC Index, SECON Index, and MMCI Index

Variables	Hospital Admission	All-Cause Mortality
	HR (95% CI)	HR (95% CI)
CoC index
≥0.7	1.00	1.00
0.4–0.7	1.61 (1.40–1.86)	1.24 (1.07–1.43)
<0.4	1.65 (1.38–1.97)	1.26 (1.05–1.51)
UPC index
≥0.7	1.00	1.00
0.4–0.7	1.43 (1.23–1.66)	1.27 (1.09–1.47)
<0.4	4.12 (2.47–6.85)	2.10 (1.13–3.91)
SECON index
≥0.7	1.00	1.00
0.4–0.7	1.55 (1.33–1.80)	1.04 (0.88–1.23)
<0.4	1.97 (1.46–2.67)	1.32 (0.95–1.83)
MMCI index
≥0.7	1.00	1.00
0.4–0.7	1.44 (1.21–1.70)	1.16 (0.98–1.38)
<0.4	2.06 (1.24–3.44)	1.13 (0.62–2.05)

Notes: Adjusted hazard ratios were estimated using Cox proportional hazards regression models. All models were adjusted for sex, age, region, income, type of medical insurance, disability status, CCI, COPD severity, number of outpatient visits, hospital location, and hospital classification.

Abbreviations: HR, Hazard Ratio; CI, Confidence Interval; MMCI, Modified Continuity Index; SECON, Sequential Continuity Index; UPC, Usual Provider Continuity.

To further characterize patterns of hospital utilization, we examined whether admissions occurred through the emergency room or via general (non-emergency) routes according to the CoC index (Table 5). Compared with patients in the high CoC group, those in the low CoC group had a significantly increased risk of general hospital admission (HR 1.70, 95% CI 1.47–1.95). A similar association was observed for hospital admission via the emergency room, with the low CoC group showing a 48% higher risk (HR 1.48, 95% CI 1.19–1.83).

Table 5.

Continuity of Care and 3-year COPD Hospitalizations by Admission Route

Variables	N (%)	CoC
		≥0.7	<0.7
		HR	HR (95% CI)
Hospital admission route
General hospital admission	1214 (68.5%)	1.00	1.70 (1.47–1.95)
Hospital admission via emergency room	559 (31.5%)	1.00	1.48 (1.19–1.83)

Notes: Adjusted hazard ratios were estimated using Cox proportional hazards regression models. All models were adjusted for sex, age, region, income, type of medical insurance, disability status, CCI, COPD severity, number of outpatient visits, hospital location, and hospital classification.

Abbreviations: HR, Hazard Ratio; CI, Confidence Interval; CoC, Continuity of Care; COPD, Chronic Obstructive Pulmonary Disease.

Discussion

This study found that lower CoC levels were associated with significantly higher risks of hospital admission and all-cause mortality among patients with COPD, and these associations were consistent across alternative continuity measures (UPC, SECON, and MMCI), admission types (general vs emergency), and most patient subgroups. The graded increase in risk with decreasing CoC, UPC, and SECON levels supports a dose–response relationship, suggesting that continuity of ambulatory care is an important determinant of both acute utilization and survival in this population.

Higher continuity is likely beneficial through several mechanisms. A stable relationship with a usual provider facilitates early recognition of symptom deterioration, proactive adjustment of maintenance therapy, and timely management of exacerbations, thereby reducing progression to hospital admission.^18,24 We found that low CoC levels were associated with general hospital and emergency room admissions, which aligns with prior research linking disrupted primary care continuity to unscheduled, potentially preventable acute care in COPD.^25,26

Subgroup analyses provided additional insight into populations in which the adverse effects of low CoC may be particularly pronounced. For hospital admission, the increased risk associated with low CoC levels was especially evident among patients primarily managed in clinics and those receiving care in rural areas. These findings suggest that CoC may act as a protective buffer in healthcare settings where structural vulnerabilities exist.

The particular association observed among patients primarily managed in clinics warrants careful consideration. In the Korean healthcare system, clinics serve as the principal setting for long-term chronic disease management, including monitoring, medication management, and timely referral.^27–29 The more than twofold increase in hospitalization risk among patients with low CoC levels indicates that fragmented outpatient care may be associated with greater healthcare utilization and worse patient outcomes.

However, these findings should be interpreted as associations rather than evidence of causality. Patients with greater disease severity may be more likely to receive care from multiple providers while also being at higher risk of hospitalization. In addition, unmeasured factors unavailable in claims data, such as smoking status, lung function, symptom burden, and healthcare-seeking behaviors, may have contributed to the observed associations. Similar mechanisms may explain the association observed in rural areas, where CoC may help patients navigate barriers to accessing healthcare services. Therefore, residual confounding cannot be excluded, and further research is needed to clarify the underlying mechanisms.

For mortality, the adverse effect of low CoC level was more evident among patients residing in metropolitan areas than among those living in non-metropolitan regions. One possible explanation is that the greater availability of healthcare providers and institutions in metropolitan areas may increase the fragmentation of care and provider switching, potentially amplifying the consequences of poor continuity. In contrast, patients in non-metropolitan areas may rely on a smaller number of providers, resulting in relatively more stable patterns of care. Nevertheless, because information regarding provider-switching behaviors, care coordination, and social support was unavailable in the present study, the mechanisms underlying this regional difference warrant further investigation.

Sensitivity analyses using alternative definitions of continuity supported the robustness of the main results. When CoC was categorized into three levels, the risks of hospital admission and death increased progressively from the highest to the lowest continuity group, indicating a clear gradient rather than a threshold-only effect. Analyses using the UPC, SECON, and MMCI indices—each capturing different aspects of continuity—also showed higher HRs in the intermediate and, in particular, the lowest continuity categories, while the direction and statistical significance of the associations remained consistent with the CoC-based findings.

Given that the CoC index incorporates both concentration and dispersion of visits, SECON reflects the stability of provider sequence over time, and UPC and MMCI highlight situations in which visits are either weakly concentrated or highly dispersed across many providers; moreover, the concordant results across these conceptually distinct metrics suggest that the underlying phenomenon is not dependent on any single metric but indicates a general lack of relational and informational continuity.³⁰ Stronger effects observed for indices that emphasize particularly low concentration or high dispersion of visits (notably UPC and MMCI) imply that outcomes are especially poor when patients have no clear usual provider or when care is fragmented across numerous, weakly connected providers.³¹ In contrast, higher CoC and SECON levels indicate that a small, relatively stable group of providers is involved in a patient’s care, potentially maintaining the protective benefit of continuity even when care is handled within a team rather than a single physician.^32,33

These findings have several policy implications. First, they reinforce continuity of ambulatory care as a modifiable target within systems that, similar to South Korea’s, allow free provider choice and often incentivize fragmented, visit-based activity. Payment and organizational reforms that promote empanelment to usual providers, continuity-oriented quality indicators, and blended or capitated payment models could help create conditions in which longitudinal relationships are rewarded rather than penalized. Second, the subgroup results highlight the need for targeted interventions in high-risk groups, such as medical-aid beneficiaries, rural residents, and clinic-managed patients, where strengthening continuity through case management, structured follow-up, shared care protocols, and regional networks may yield disproportionate reductions in hospital admissions and mortality. Finally, because continuity proved important across multiple measurement approaches and admission types, incorporating continuity metrics (such as CoC ≥0.7) into national quality monitoring and chronic disease programs may help align clinical practice, health-system design, and patient outcomes toward more coordinated, person-centered care for COPD.³⁴

This study has several limitations. First, the analysis was based on administrative claims data, which lack detailed clinical information such as lung function parameters, smoking status, symptom burden, and health behaviors. Consequently, residual confounding by unmeasured factors cannot be excluded. In addition, although all-cause mortality is a robust and clinically meaningful outcome, the claims database did not provide sufficiently detailed information to distinguish respiratory-related deaths from non-respiratory causes, such as cardiovascular mortality. Therefore, we were unable to evaluate whether the observed association between CoC and mortality differed according to specific causes of death, limiting mechanistic interpretation of the findings. Second, restricting the study population to patients with at least three outpatient visits during the 1-year baseline period may have introduced selection bias by excluding individuals with limited healthcare utilization or rapidly progressive disease. Therefore, the findings may not be fully generalizable to patients with early fatal disease trajectories. Third, the CoC, UPC, SECON, and MMCI indices were calculated from visit patterns observed during a fixed period and may not fully capture relational aspects of continuity, such as perceived trust, communication quality, or informal contacts outside reimbursed visits. Fourth, this study was conducted within the Korean healthcare system, which is characterized by relatively unrestricted access to specialists and hospitals without a strong gatekeeping mechanism. Therefore, the findings may be most applicable to healthcare systems with similar access patterns, and caution is warranted when generalizing the results to settings with different referral pathways or payment structures. Finally, although subgroup and sensitivity analyses yielded generally consistent findings, multiple comparisons increase the possibility of chance findings, and some subgroup estimates were based on relatively small numbers of events. Future prospective studies incorporating detailed clinical information and patient-reported measures of continuity are needed to further clarify the mechanisms underlying these associations.

Despite these limitations, this study utilized a large, nationwide claims database from a single-payer insurance system, allowing inclusion of a real-world COPD population with high statistical power and excellent representation of hospital admission and mortality rates. The use of routinely collected data also minimizes selection bias related to healthcare access or specific institutions. Second, continuity was quantified with four complementary, validated indices (CoC, UPC, SECON, and MMCI), and the main findings were robust across multiple categorizations and sensitivity analyses. This dual-measure approach reinforces construct validity and reduces concerns that the observed associations are artefacts of a particular continuity metric or threshold. Third, the study simultaneously examined hospital admission and all-cause mortality, as well as types of admission (general vs emergency room), providing a comprehensive understanding of how continuity influences multiple clinically meaningful outcomes in COPD. The consistent trend in these associations across outcomes supports a coherent underlying mechanism linking fragmented care to worse clinical trajectories. Finally, detailed subgroup analyses by socioeconomic status, region, healthcare setting, comorbidity burden, and COPD severity identified patient groups in whom continuity appears especially important, such as medical-aid beneficiaries, rural residents, clinic-managed patients, and those with severe disease. These findings may enhance the clinical and policy relevance of the work by highlighting priority populations for continuity-enhancing interventions.

Conclusion

In this nationwide cohort of patients with COPD, lower CoC level was consistently associated with higher risks of hospital admission and all-cause mortality. Similar trends were observed across UPC, SECON, and MMCI categories. Overall, these associations were robust in multiple subgroup and sensitivity analyses, indicating that fragmented outpatient care is closely linked to worse outcomes irrespective of continuity assessment method or hospitalization type. While causal relationships cannot be established from this observational study, the findings underscore the potential importance of CoC in COPD management and support the inclusion of patient-level continuity indicators in future quality assessment and chronic disease management strategies.

Funding Statement

No funding was received for this study.

Abbreviations

CoC, Continuity of Care; COPD, Chronic Obstructive Pulmonary Disease; NHIS, National Health Insurance Service; UPC, Usual Provider of Care; SECON, Sequential Continuity; MMCI, Modified Modified Continuity Index; HR, Hazard Ratio; CI, Confidence Interval; ICD-10, International Classification of Diseases 10th Revision.

Data Sharing Statement

The data supporting the findings of this study are available from the National Health Insurance Sharing Service; However, access to these data is restricted and subject to approval. The data is available after acceptance of approval for use by the national health insurance service (https://nhiss.nhis.or.kr/bd/ab/bdaba000eng.do).

Ethics Approval and Informed Consent

This study was reviewed and approved by the International Deliberative Committee on the Health Care System of Yonsei University (Registration No.: 4-2025-0212). The requirement for informed consent was waived by the committee because the study utilized de-identified routine administrative and billing data.

Consent for Publication

Not applicable. The study did not involve identifiable human data or images requiring consent for publication.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no competing interests in this work.