Risk factors for interstitial lung disease: a 9-year Nationwide population-based study

The National Health Insurance Service (NHI) covers more than 99% of all Korean residents and includes all health claim data, including diagnostic codes, procedures, prescription drugs, patient personal information, and hospital information. There is one health insurance system with a unique resident registration number for each citizen; therefore, duplication of subjects can be avoided. This study used data from the National Health Insurance Service-National Sample Cohort(NHIS-NSC) 2002–2013 [9], which was released by the KNHIS in 2015. It includes all medical claims filed from January 2002 to December 2013 for 1,099,094 nationally representative randomly selected subjects, accounting for approximately 2.2% of the entire population in the KNHIS in 2002. The data were produced by the KNHIS using a systematic sampling method to generate a representative sample of all 46,605,433 Korean residents in 2002. The cohort population underwent biennial medical evaluations through the NHI Corporation between January 1, 2002, and December 31, 2013.

This was a nationwide population-based 9-year longitudinal study. We included subjects with a self-reported smoking history, who were ≥ 40 years old, and with co-morbidities diagnosed before the index date. Health examination data confirmed self-reported cigarette smoking history. The health examination data every 2 years was linked with the cohort data.

We did not include patients with ILD from 2002 to 2004 to exclude preexisting cases of ILD during the two medical evaluations and 1 year of follow-up. We also excluded patients with ILD who did not visit the clinic 30 days after the index date. The final sample included 312,519 subjects. Then, we identified patients with newly diagnosed ILD between January 2005 and December 2013. Among the 312,519 subjects, the control group was selected by excluding those who had ILD between 2005 and 2013.

Cases of ILD were selected based on International Classification of Disease-10 (ICD-10) code J84 for other interstitial lung diseases, excluding drug-induced interstitial lung disorders, interstitial emphysema, and lung diseases caused by external agents. Connective tissue disease-associated ILD, hypersensitivity pneumonitis, and sarcoidosis were excluded.

Comorbidities diagnosed before the index date that could be associated with an increased risk of lung fibrosis were identified using ICD-10, including COPD, hepatitis C, gastroesophageal reflux disorder (GERD), and diabetes [10‐13].

The final sample included 312,519 subjects, of which 1972 developed ILD during the 9-year study period (Fig. 1). ILD incidence was 70.1 cases per 100,000 person-year.

All subjects were tracked by December 31, 2013, Follow up duration was median 65.6 months (interquartile range: 35.5, 89.0) in the ILD group and median 107.5 months (interquartile range; 100.2, 109.5) in the non-ILD group.

The ILD group had a higher percentage of men than the control. Compared with the control, subjects with ILD were older, and more likely to be smokers. The ILD group was more likely to have comorbidities such as respiratory diseases (tuberculosis and pneumonia), diabetes, chronic renal failure, malignancy, GERD, hepatitis C, and COPD than the control (Table 1).

Based on a multivariate Cox regression analysis of all variables (Table 2), smoking was significantly associated with the development of ILD (HR: 1.2; 95% CI: 1.1–1.4).

Co-morbidities such as hepatitis C (HR: 1.6; 95% CI: 1.1–2.3), history of tuberculosis (HR: 1.5; 95% CI: 1.1–1.9), history of pneumonia (HR: 1.6; 95% CI: 1.3–2.0), and COPD (HR: 1.8; 95% CI: 1.6–2.1) were significantly associated with the development of ILD. Men were almost twice as likely to associated with the development of ILD as women.

The risk of ILD development was 1.9 times in the 50s, 4.1 times in the 60s, and 6.9 times in the 70s, which increased sharply with age. In multivariate analysis, age was found to be most associated to ILD development showing dose response. All the risk factors included in the model showed relatively narrow confidence intervals.

There were no variables that changed the direction of the hazard ratio in the univariate and multivariate analyses except hepatitis B, which was statistically not significant. The hazard ratios of COPD and history of tuberculosis were reduced by almost half after multivariate analysis. History of pneumonia, hepatitis C and diabetes showed also reduced hazard ratios in multivariate analysis. Smoking was associated with the occurrence of ILD, but its magnitude was relatively small compared to other variables.

Multivariate analysis stratified by smoking showed same direction and similar magnitude of hazard ratios in all variables. However, hazard ratios stratified by sex were somewhat different in several variables. As a result of stratified analysis on the basis of gender, the hazard ratios of COPD according to man and woman were similar 1.8 and 1.9, respectively. In the case of GERD, multivariate analysis of male subjects showed that GERD was not a significant risk factor, but the hazard ratio for women was 1.3, which was consistent with the 1.3 observed in the univariate analysis (Table 3). GERD, Hepatitis C are significant risk factors in females, in contrast malignancy and diabetes are significant risk factors in males.

The most important limitation of the present study is that the diagnoses of ILD and other comorbidities were defined based on ICD codes, which may be inaccurate compared to the diagnoses obtained from a medical chart. Underreporting of asymptomatic ILD or misclassification was also possible.

The validity of the medical insurance claims data for ILD has not been determined in Korea. This database consists of random samples of national insurance claim data without identification numbers. Therefore, it was impossible to validate individual cases through a chart review.

We may have underestimated ILD incidence because of inaccurate ILD data. However, previous studies of ILD incidence using data from the Health Insurance Review and Assessment Service of Korea [38] reported similar results to those of previous studies [35, 36, 39]. The incidence rate of ILD was reportedly 48.5 per 100,000 person-years in Korea based on all claims data from 2008 to 2012 [38]. In the present study, the incidence rate was 70.1 per 100,000 patients from 2005 to 2013, which appears reasonable because we excluded subjects aged < 40 years.

The present study did not take into consideration of relevant risk factor such as occupational and environmental exposure. However, annual income could be a weak proxy of the occupational exposure [40].

The present study may be affected by selection bias because the controls were identified based on medical claims. Thus, the controls were more likely to have comorbidities than controls selected from the general population. Although we excluded patients with a diagnosis of ILD between January 2002 and December 2004 to exclude pre-existing ILD diagnosed before the first year of our study (2005), patients with ILD could be miscounted as new cases if the patient did not require medical care between 2002 and 2004, which may have caused inaccuracies. In addition, other risk factors, such as pulmonary function and high-resolution computed tomography findings for ILD, could not be evaluated because of the NHIS-NSC 2002–2013 primarily included medical claims.