Treatment patterns, complications, and direct medical costs associated with ankylosing spondylitis in Chinese urban patients: a retrospective claims dataset analysis

Abstract

Aim: To describe treatment pattern, complications, and direct medical costs associated with ankylosing spondylitis (AS) in Chinese urban patients.

Methods: The 2013 China Health Insurance Research Association (CHIRA) urban insurance claims database was used to identify patients with AS. The identified patients were stratified by AS treatments for the comparisons of well established AS-related complications and direct medical costs. Conventional regression analyses adjusted the collected patient baseline characteristics to confirm the impact of treatments on complications and direct medical costs.

Results: Of the identified 1299 patients with AS, 18.0% received non-steroidal anti-inflammatory drugs (NSAID), 11.2% received immunosuppressant, 48.2% received NSAID plus immunosuppressant, 4.6% received biologic agents, and 17.9% received medications without indication for AS. Biologic group was associated with the lowest proportion of AS-related complications (8.3%) that was confirmed by multiple logistic regression analysis (odds ratio = 0.200, p = .017). The biologic group was also associated with highest direct medical costs (median: RMB = 14,539) that were confirmed by the multiple generalized linear model (coefficient = 1.644, p < .001).

Conclusions: Biologics were not commonly used for AS in Chinese patients likely due to their high cost. Future studies are needed to confirm the potential long-term clinical benefits associated biologic treatment for AS.

Keywords:

• Ankylosing spondylitis
• Treatment pattern
• Direct medical costs
• Complications
• China

Introduction

The prevalence of ankylosing spondylitis (AS) in China is 0.3%¹. The main clinical symptoms associated with AS are long-term pain and stiffness of spine². The persistent AS could lead to bone spur with ligament and bamboo-like spine that significantly limit patient physical function, impair patient quality-of-life, and reduce patient productivity³. Thus, the primary treatment goal for AS is to control the AS-related symptoms and delay the disease progression⁴.

Biologic agents targeting tumor necrosis factor-α (TNF-α), the chimeric monoclonal IgG1 antibody infliximab, and the recombinant 75-kD TNF receptor IgG1 fusion protein etanercept, have substantially changed the AS treatment landscape in high-income countries^5,6, because of their superior treatment effects over conventional AS treatments, such as non-steroidal anti-inflammatory drugs (NSAID) and immunosuppressant^7,8. Biologics have been used for AS in China for over 10 years. However, the impact of biologic agents on the treatment pattern and disease burden in Chinese patients with AS hasn’t been well studied due to limited data availabilities and access. Since 2013, the China Health Insurance Research Association (CHIRA) started to use a longitudinal approach to create a national database consisting of claims records of randomly selected urban patients across China. Thus, this database enables studies describing the treatment pattern, estimating the prevalence of disease-related complications, and measuring direct medical costs associated with Chinese patients living in urban areas.

Patients and methods

The 2013 CHIRA claims database randomly sampled 6.25 million claimants of urban workers and residents public insurance plans in 25 provinces and four central cities of China⁹. As of 2013, the urban workers and residents insurance plans covered over 90% of the urban population in China¹⁰. Because the 2013 CHIRA claims database only contained the 1-year claims records associated with selected claimants in 2013, our study was designed as a cross-sectional study using the 2013 CHIRA claims database to describe treatments for AS, AS-related complications, and direct medical costs in Chinese urban patients with AS in a 1-year period from January 1, 2013 to December 31, 2013.

Data sources

The 2013 CHIRA database consisted of two datasets, the general dataset and the claims record dataset. The general dataset contained patient demographics, insurance plan, hospital admissions, the Tenth revision of the international classification of disease (ICD-10) diagnosis codes associated with hospital admissions and outpatient clinic visits. The general dataset also included aggregated direct medical costs classified as drug costs and non-drug costs. The claims record dataset contained each claims record associated with in-patient care and out-patient care. Each claims record contained health resource name, transaction date, health resource unit cost, and purchased health resource units. The two datasets can be linked through the unique patient identification number.

Patient identification and data extraction

The ICD-10 diagnose code for AS (M45) was used as the keyword to search AS patients in the general dataset. The extracted information associated with identified patients with AS included patient demographics, insurance plans, patient residence city, hospital admissions, diagnosis codes, aggregated drug costs, and aggregated non-drug costs. Even though the current Chinese public health insurance plans, including urban workers and residents plans, don’t make full reimbursement of utilized health resources, the utilized health resources were recorded in the insurance claims database, irrespective of their reimbursement status. The identified patients were further linked with the claims record dataset to extract their prescribed medications. The process of patient identification and data extraction from the 2013 CHIRA claims database is illustrated in Figure 1.

Figure 1. The process of patient identification and data extraction from the 2013 CHIRA claims database to describe treatment pattern, complications, and direct medical costs associated with AS in Chinese patients.

Outcome measures

Our study collected the following outcomes. First, the identified diagnosis codes associated with hospitalizations and outpatient clinic visits were used to assess the AS-related complications, which included iridocyclitis (ICD-10 code: H20), coxarthropathy (ICD-10 code: M12.05), lung diseases including sarcoidosis (ICD-10 code: D86) and bronchiectasis (ICD-10 code: J47), and chronic renal disease (ICD-10 code: N18)¹¹. Second, we classified the prescribed medications as medications without indication for AS (non-AS treatment), non-steroidal anti-inflammatory drugs (NSAIDs), immunosuppressant, and biologics according to AS treatment guidelines⁴. Third, the health resources utilization were aggregated to calculate the number of hospital admissions, the length of hospital stay, and the number of outpatient clinic visits in 2013. Finally, the direct medical costs associated with the included patients with AS were aggregated as the sum of drug costs and non-drug costs, respectively. All costs in this study were presented with 2013 Chinese currency Renminbi (RMB), which could be exchanged for US dollar at the rate of 1:0.16.

Statistical data analysis

The treatment pattern was described using the distribution of the medications that were classified as non-AS treatment, NSAID, immunosuppressant, NSAID plus immunosuppressant, and biologics. The classified medications were further used to stratify the included patients for unadjusted comparisons of patient baseline characteristics, the prevalence of AS-related complications, health resources utilization, and direct medical costs. The statistical methods used for the unadjusted comparisons between non-AS treatment group and the other four AS treatment groups included Student t-test for continuous variables/outcomes, Fisher’s exact test for dichotomous variables/outcomes, Poisson distribution test for count outcomes including hospital admissions and out-patient visits, and Wilcoxon rank sum test for cost outcomes. Multiple logistic regression analysis and generalized linear regression analyses with adjustment for patient baseline characteristics and the classified treatments were conducted to confirm the statistical significance detected in the unadjusted comparisons. The statistical significance defined in this study was two-sided p-value less than 0.05. SAS 9.2 was used to perform the statistical data analyses.

Results

The search of the general dataset of the 2013 CHIRA database identified 1,299 patients with AS diagnosis code. The prescribed medications in these identified patients included NSAID plus immunosuppressant (48.2%), NSAID (18.0%), non-AS treatment (17.9%), immunosuppressant (11.2%), and biologics (4.6%).

Treatment pattern and patient baseline characteristics

Among the prescribed medications in the identified 1,299 AS patients, diclofenac was the most frequently used NSAID (29.1%), salazosulfapyridine was the most commonly used immunosuppressant (30.2%), and the recombinant human tumor necrosis factor-Fc (Etanercept) was the most often used biologic agent (3.6%). When compared to the non-AS treatment group (patients receiving medications without indication for AS), the NSAID group was associated with significantly higher proportions of patients living in the east of China (73.1% vs 64.4%, p = .046), patients visiting tier III hospitals (86.8% vs 72.5%, p < .001), and patients with comorbidities for sacroiliac disease (18.4% vs 7.3%, p < .001), lung diseases (18.4% vs 6.0%, p < .001), hypertension (14.5% vs 3.4%, p < .001), gastritis (12.4% vs 5.6%, p = .014), osteoporosis (6.0% vs 1.7%, p = .028), and osteoarthritis (8.1% vs 0.4%, p < .001). The immunosuppressant group and the non-AS treatment group had highly comparable patient baseline characteristics except the prevalence of osteoarthritis (4.8% vs 0.4%, p = .006). The NSAID plus immunosuppressant group and the NSAID group had similar patient baseline characteristics. Finally, the biologic group differed from the non-AS treatment group by including more patients living in central cities (38.3% vs 17.2%, p = .001), more patients visiting higher rank tertiary hospitals (visiting tier III hospitals: 95.0% vs 72.5%, p < .001), and more patients with rheumatoid arthritis (15.0% vs 0.4%, p < .001). The patient baseline characteristics associated with the five treatment groups are summarized in Table 1.

Table 1. Comparisons of patient baseline characteristics associated with the identified patients with AS between the non-AS treatment group and the other four AS-related treatment groups.

Variable	Non-AS treatment	NSAID		Immunosuppressant		NSAID plus immunosuppressant		Biologics
	Mean (%)	Mean (%)	p value	Mean (%)	p value	Mean (%)	p value	Mean (%)	p value
Sample size	233	234		146		626		60
Distribution, %	17.9	18.0		11.2		48.2		4.6
Age, years	40.3 ± 13.4	42.5 ± 14.4	.086	39.0 ± 11.2	.345	40.81 ± 13.4	.624	37.6 ± 12.7	.184
Male, %	57.5	65.0	.107	67.1	.066	68.7	.003	71.7	.054
Geographic location, %
East	64.4	73.1	.046	61.6	.661	74.9	.003	61.7	.764
Middle	20.2	13.7	.065	23.3	.52	17.1	.317	21.7	.858
West	15.5	13.3	.512	15.1	1	8.0	.002	16.7	.843
Residence city
Provincial capital city	48.5	47.0	.781	56.2	.17	52.4	.319	56.7	.311
Central city	17.2	21.8	.243	12.3	.241	19.7	.435	38.3	.001
Regional city	34.3	31.2	.491	31.5	.578	28.0	.078	5.0	<.001
Hospital setting
Tier I	35.2	41.0	.216	38.4	.584	31.2	.286	30.0	>.542
Tier II	21.9	23.9	.66	21.9	1	23.5	.65	23.3	>.862
Tier III	72.5	86.8	<.001	80.1	.111	96.2	<.001	95.0	<.001
Insurance type
Worker	88.4	89.7	.659	93.2	.156	90.1	.453	86.7	.661
Resident	11.6	10.3	.659	6.9	.156	9.9	.453	13.3	.661
AS-related complication
Iridocyclitis	0.0	2.1	.061	1.4	.148	3.4	.002	1.7	.205
Coxarthropathy	7.3	18.4	<.001	11.0	.262	18.7	<.001	6.7	1
Back pain	1.3	4.3	.088	2.1	.68	3.0	.223	1.7	1
Lung disease	6.0	18.4	<.001	9.6	.227	14.9	<.001	1.7	.321
Renal disease	1.7	2.6	.751	1.4	1	1.3	.744	0.0	.585
Any complications	13.3	30.3	<.001	17.1	.372	25.1	<.001	8.3	.381
Comorbidities
Hypertension	3.4	14.5	<.001	4.8	.591	13.3	<.001	1.7	.691
Coronary heart disease	0.9	5.6	.004	0.7	.574	5.1	.004	0.0	.351
Diabetes	3.4	8.1	.046	2.1	.541	5.6	.222	0.0	.367
Gastritis	5.6	12.4	.014	6.9	.661	8.3	.195	1.7	.314
Osteoporosis	1.7	6.0	.028	2.1	1	6.7	.003	0.0	.585
Rheumatoid arthritis	0.4	2.6	.122	2.1	.161	6.2	<.001	15.0	<.001
Osteoarthritis	0.4	8.1	<.001	4.8	.006	10.5	<.001	3.3	.108

AS-related complications

Of the stratified five treatment groups, the biologic group was associated with the lowest proportion prevalence of any AS-related complications (8.3%). Further adjustment of collected patient baseline characteristics through multiple logistic regression analysis still observed statistical significance for the lowest odds ratio (OR) (0.200, p = .017) associated with biologics for any AS-related complications among the five classified treatments. Additionally, the multiple logistic regression analysis observed significantly higher risk of any AS-related complications associated with NSAID (OR = 2.327, p = .008), NSAID plus immunosuppressant (OR = 1.735, p = .05), age (OR = 1.049, p < .001), patients living in the east of China (OR = 3.799, p < .001) and central cities (OR = 9.373, p < .001), and the comorbidities with gastritis (OR = 8.493, p < .001) and osteoarthritis (OR = 252.464, p < .001) (Figure 2).

Figure 2. The results of multiple logistic regression analysis assessing the association between AS-related treatments and AS-related complications in the identified patients with AS.

Health resources utilization and direct medical costs

When using the non-AS treatment group as the reference, NSAID, NSAID plus immunosuppressant, and biologic treatment groups were associated with significantly more utilization of health resources. For instance, significantly more outpatient clinic visits were observed in the NSAID group (19.8 visits vs 8.0 visits, p < .001) and NSAID plus immunosuppressant group (20.5 visits vs 8.0 visits, p < .001). Significantly more hospital admissions and longer length of hospital stay were observed in the NSAID plus immunosuppressant group (hospital admissions: 1.1 vs 0.6, p < .001; hospital stay length: 13.9 days vs 8.3 days, p < .001) and the biologic group (hospital admissions: 2.7 vs 0.6, p < .001; hospital stay length: 21.5 days vs 8.3 days, p < .001), respectively. The direct medical costs of five treatment groups ranged from RMB2,702 for the non-AS treatment group to RMB 14,539 for the biologic treatment group (Table 2). Multiple generalized linear regression analyses with full adjustment of collected patient baseline characteristics confirmed that the association between biologics and direct medical costs was ranked as the highest among the five treatments (coefficient = 1.644, p < .001) (Table 3).

Table 2. Health resources utilization and direct medical costs associated with the defined treatment groups from the identified patients with AS.

Treatment group	Non-AS treatment		NSAID			Immunosuppressant			NSAID plus immunosuppressant			Biologics
Outcomes	Mean	SD	Mean	SD	p value	Mean	SD	p value	Mean	SD	p value	Mean	SD	p value
Sample size	233		234			146			626			60
Health resources utilization
Out-patient clinic visits	8.0	18.4	19.8	34.4	<.001	8.7	16.5	.718	20.5	34.0	<.001	13.0	27.1	.094
Hospital admissions	0.6	0.7	0.7	0.9	.371	0.6	0.8	.794	1.1	1.2	<.001	2.7	2.8	<.001
Hospital stay length	8.3	12.3	10.8	16.5	.067	7.9	13.5	.774	13.9	18.4	<.001	21.5	21.0	<.001
Direct medical costs	Median	(Min, Max)	Median	(Min, Max)	p value	Median	(Min, Max)	p value	Median	(Min, Max)	p value	Median	(Min, Max)	p value
Drug costs	748	(0, 30,171)	2,606	(27, 65,712)	<.001	1,048	(15, 20,001)	.048	3,837	(37, 78,137)	<.001	9,895	(1,154, 99,950)	<.001
Non-drug costs	968	(0, 58,907)	2,365	(0, 64,581)	<.001	1,960	(0, 52,846)	.167	3,669	(0, 1,31,713)	<.001	4,695	(0, 35,993)	<.001
Total costs	2,702	(2, 62,683)	5,572	(33, 89,739	<.001	3,900	(72. 56,336)	.042	8,055	(37, 15,0611)	<.001	14,539	(2,932, 11,65,733)	<.001

Table 3. The summary of the results from the multiple generalized linear regression analyses assessing the impact of AS-related treatments on direct medical costs in the identified 1299 patients with AS.

Variable name	Drug costs			Non-drug costs					Total direct medical costs
	Coefficient	95% CI		p value	Coefficient	95% CI		p value	Coefficient	95% CI		p value
		Lower limit	Upper limit			Lower limit	Upper limit			Lower limit	Upper limit
Age	0.014	0.009	0.019	<.001	0.016	0.011	0.022	<.001	0.015	0.010	0.020	<.001
Hospital setting
Tier III hospital	0.708	0.501	0.915	<.001	0.220	−0.040	0.480	.097	0.553	0.348	0.758	<.001
AS-related complication
Spinal fracture	0.398	0.441	1.237	.353	1.405	0.525	2.284	.002	1.071	0.237	1.905	.012
Coxarthropathy	−0.168	−0.467	0.131	.271	−0.131	−0.452	0.190	.424	−0.135	−0.431	0.162	.373
Comorbidities
Hypertension	0.552	0.322	0.782	<.001	−0.117	−0.371	0.137	.366	0.260	0.029	0.491	.027
Gastritis	0.149	−0.102	0.400	.246	−0.362	−0.645	−0.078	.012	−0.071	−0.325	0.183	.586
Osteoporosis	0.543	0.206	0.881	.002	0.527	0.150	0.903	.006	0.485	0.143	0.826	.005
AS-related treatment
Non-AS treatment	0.000	0.000	0.000	–	0.000	0.000	0.000	–	0.000	0.000	0.000	–
NSAID	0.346	0.127	0.565	.002	0.320	0.076	0.564	.010	0.400	0.190	0.611	<.001
Immunosuppressant	−0.116	−0.357	0.124	.343	0.247	−0.037	0.531	.088	0.082	−0.153	0.317	.495
NSAID plus immunosuppressant	0.596	0.406	0.786	<.001	0.765	0.553	0.977	<.001	0.782	0.599	0.965	<.001
Biologics	1.961	1.626	2.297	<.001	0.849	0.493	1.206	<.001	1.644	1.316	1.973	<.001

Bold indicate statistical significance defined as p value < 0.05

Discussion

The 2013 CHIRA database includes the reimbursement records of over 6 million urban workers and residents across China and enables us to identify a relatively large cohort of AS patients. The treatment pattern of AS suggested that the introduction of biologic agents for AS had little impact on the AS treatment landscapes in China and the current mainstream therapy for AS is the combination of NSAID and immunosuppressant, which was also reported by a treatment pattern survey study for rheumatoid arthritis in China¹². Our study found the lowest prevalence of AS-related complications in patients receiving biologics. Even though the impact of biologics on the clinical symptoms and disease activity of AS has been well established, the long-term clinical benefits, such as the reduction of complications, are seldom reported because of slow disease progression. Thus, this study finding could be used as preliminary evidence encouraging future studies to confirm the potential long-term clinical benefits associated with biologics for AS. As expected, biologic treatment associated with much higher drug costs than conventional treatment because of high acquisition costs. However, the higher non-drug costs and more hospital admissions related to biologics were inconsistent with the observed impact of biologics on health resources utilization associated with AS in high-income countries^13,14.

The treatment pattern for AS in our study cohort indicated a significant barrier for patient access to biologic agents in China. Except the barrier associated with higher acquisition costs of biologics, there could be some other factors discouraging the utilization of biologics for AS in China. First, AS is a chronic disease that could last for decades without a noticeable short-term survival impact on patients. The medical needs of AS patients are not well recognized and the new treatments for AS are unlikely to be taken into account as the reimbursement priority by the Chinese public payers. Second, there’s insufficient robust local evidence confirming the real world effectiveness of biologics. Third, the typical management settings for AS are community hospitals, where the knowledge translation for new treatment usually takes a much longer time¹⁵. Thus, patients in our study received biologics treatment mainly through Tier III hospitals, the highest rank hospitals in China. Finally, the hospital-based reimbursement policy in China could be another significant barrier to the access to biologics treatment, as patients might reduce their willingness to use biologics because of the time-consuming process of hospitalization¹⁶.

Our study observed a significantly higher number of hospital admissions and longer length of hospital stay in patients receiving biologic treatment, which was against our hypothesis that biologics could reduce hospitalizations by improving clinical symptoms and reducing disease activity. According to the discussion with the clinical rheumatologists about this finding, patients usually receive biologic treatment in hospital settings under the current reimbursement policy which only covers hospital drugs. The present hospital-based reimbursement policy could increase hospitalizations in patients receiving biologics treatment and offset the potential cost savings associated with treatment effects of biologic therapies. Thus, our study could be a case study to demonstrate the confounding effects related to reimbursement policy for health resources utilization. The confounding effects associated with the reimbursement policy in patients receiving biologics are difficult to be adjusted in our study.

The observed lowest prevalence of AS-related complications in patients receiving biologic treatment should not be over-interpreted due to the limitations associated with our data source, which doesn’t include sufficient information to verify the identified AS-related complications and the onset of biologic treatments. However, the cumulated clinical evidence in other studies suggest that biologic treatments could improve long-term health outcomes in patients with AS. For example, one study compared the radiographic progression of AS patients treated with infliximab vs historical controls never treated with biologics over 8 years and observed a significantly lower number of new syndesmophytes in patients receiving infliximab (1.0 ± 0.6 vs 2.7 ± 0.8, p = .007)¹⁷. Another registry study following up 1,436 patients receiving biologics TNF-α inhibitor observed better survival in those patients without treatment switching¹⁸. Because the lack of response was the primary reason for treatment switching, the sustainable response to biologics could reduce the risk of AS-related complications that have a substantial impact on patient survival. Thus, our finding of the potential association between biologics and reduced risk of AS-related complications could be used as preliminary evidence encouraging future studies to confirm the long-term clinical benefits associated with biologics for AS in Chinese patients and generate more robust evidence to support the use of biologics for AS.

As the only national representative database for urban workers and residents, the 2013 CHIRA database identified a considerable sample size to give a reliable description of treatment pattern, health resources utilization, and direct medical costs for AS in China. However, cautions to the primary limitations associated with the database are still needed when interpreting our study findings. First, this database collected very minimum clinical information related to hospitalization. The collected information from this data source was unable to track previous treatment history, estimate the actual duration of biologic therapies, and assess disease activity. Thus, the adjustments in our regression analyses are highly limited, and the unknown confounding effects could substantially bias our study results. Second, the observation time in this data source was only 1 year, which was unlikely to be long enough to observe the impact of AS-related treatments on long-term health outcomes and health resources utilization. Third, the 2013 CHIRA database didn’t collect utilization information of medications that were purchased outside hospital settings. This missing information could under-estimate direct medical costs and bias the comparisons of health resources utilization and direct medical costs in our study. Because high cost medications are often paid from patient out-of-pocket costs that couldn’t be fully captured by the claims database¹⁹, the utilization of biologics in our study was likely to be under-estimated. Finally, the database doesn’t include rural patients. Thus, the evidence generated from this study has limited generalizability when used to support reimbursement decision-making across China.

Our study suggested that the use of biologics for AS in Chinese patients is uncommon. The study showed the lowest prevalence of AS-related complications associated with biologics treatment, suggesting possible long-term clinical benefits related to biologic treatments. Future studies are needed to confirm the reduced AS-related complications in patients receiving biologic therapies. The current hospital-based reimbursement policy in China could confound hospital admissions and length of stay in patients receiving biologic treatments and offset the potential cost savings associated with clinical benefits of biologics treatment.

Transparency

Declaration of funding

This manuscript was funded by Xian Janssen.

Declaration of financial/other relationships

QL and SG are employees of Beijing North Medical & Health Economics Research Center, a research organization receiving industry funds/grants for claims data analysis. YC and WC are the employees of Normin Health, a consulting firm receiving industry funds for health economics and outcomes research. YY is an employee of Xian Janssen. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.