Spatiotemporal distribution and clinical manifestations of moyamoya disease in Jiangxi Province, China: a retrospective hospital-based study

Abstract

Introduction

Moyamoya disease (MMD) has been reported to be continuously increasing in China, with Jiangxi Province having the highest incidence nationwide. However, limited by data sources, national survey results have difficulty accurately reflecting the true incidence of MMD, indicating that detailed epidemiological data for this highest incidence province remain unavailable.

Methods

We extracted medical records of MMD patients from 210 secondary and tertiary hospitals in Jiangxi Province from 2015 to July 2022. Spatiotemporal analysis was adopted to explore the spatial and temporal patterns and hospital burden of MMD in different counties and municipalities of Jiangxi Province.

Results

A total of 18 693 hospitalisation records were included in this study, of which 11 140 were new-onset patients. The average annual incidence during 2015–2022 was 3.36 per 100 000 residents, with an average annual prevalence of 5.65 per 100 000 residents. The incidence exhibited a clustered regional distribution pattern (Moran’s I=0.51, Z-score=8.14, p<0.001). The peak incidence and prevalence occurred in adults aged 51–55 years. Common comorbidities included hypertension (36.4%), diabetes (11.1%) and hyperlipidaemia (6.2%). The mean annual hospitalisation cost was approximately CNY 29,409, with basic medical insurance covering about 67% of payment methods.

Conclusions

This study is the first to comprehensively characterise the epidemiological characteristics and hospital burden of MMD in the highest incidence province of China. Our findings contribute to the development of effective public health policies, may provide novel insights into the aetiology of MMD and reduce the burden of MMD.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Based on the latest epidemiological survey data, the annual incidence rate of moyamoya disease (MMD) in China was 1.14 patients per 100 000 people from 2016 to 2018, which increased by about two times over 3 years. Notably, the incidence rate showed a spatial clustering distribution pattern, with Jiangxi Province having the highest incidence rate in the country.

WHAT THIS STUDY ADDS

• This study employs hospitalisation records and geospatial analysis techniques to investigate the epidemiological characteristics and spatiotemporal patterns of hospitalisation burden associated with MMD in Jiangxi Province.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This paper provides a comprehensive analysis of the spatiotemporal characteristics of MMD epidemiology and the associated hospitalisation burden in this region, with the aim of offering scientific evidence for the prevention and treatment of the disease. The findings of this study are of great significance for formulating effective public health policies and reducing the burden of MMD. Moreover, the results may provide new insights for future epidemiological and etiological investigations of MMD in key areas.

Introduction

Moyamoya disease (MMD) is a progressive stenosis and occlusion of the intracranial internal carotid arteries and their branches, leading to the formation of a compensatory collateral vascular network at the skull base, which predisposes affected individuals to cerebrovascular events including cerebral infarction and ischaemic stroke.¹ Recent studies have revealed that the incidence and prevalence of MMD are on the rise worldwide, especially in East Asia, which has received increasing attention and widespread concern.²

According to two early national surveys conducted in Japan, the total number of patients treated was estimated to be 3900 in 1994, with a prevalence and incidence of 3.16 and 0.35 per 100 000 people, respectively.³ By 2003, the estimated number of MMD patients in Japan had almost doubled to 7700, with a new diagnosis rate of 0.54 patients per 100 000 persons annually.⁴

Nonetheless, despite reports on the incidence of MMD in China, the comprehensiveness and accuracy of these reports are limited due to the constraints of data sources. Miao et al initially reported an annual incidence of 3.92 per 100 000 in 15 major hospitals in Nanjing from 2000 to 2007, with an average annual detection rate of 0.43 per 100 000 persons.⁵ In 2018, Bao et al provided preliminary findings on MMD in China, reporting that transient ischaemic attack was the most common initial clinical manifestation (48.13%), followed by stroke (22.62%), haemorrhage (16.45%) and headache (5.57%). The number of incident cases exhibited geographical differences, uncovering emerging perspectives and new trends in MMD in China.⁶ It should be noted that the apparent increase in MMD incidence may be partly attributable to advances in neuroimaging technology and the improvement of diagnostic criteria. Despite these methodological enhancements, a discernible uptrend in MMD incidence rates has been observed in recent years. The latest epidemiological survey data from 2016 to 2018 report an annual incidence rate of 1.14 per 100 000 individuals in China, marking an approximate twofold increase within this span. Notably, the incidence rate has demonstrated a spatial clustering distribution, with Jiangxi Province exhibiting the highest incidence levels nationwide.⁷

Jiangxi, located in southeastern China (114°02′ to 118°28′ E, 24°7′ to 29°9′ N), is renowned for its significant natural and industrial resources, including Poyang Lake, the country’s largest freshwater lake, and a wealth of mineral resources, positioning it as one of Asia’s premier copper industrial bases. The province features hilly and mountainous terrain south of the Yangtze River, with a predominantly humid subtropical climate. As of 2021, Jiangxi had an estimated population of 45.18 million, with Han Chinese comprising 99.51% of the population (online supplemental figure 1). In terms of economic indicators, Jiangxi ranks 15th among Chinese provinces in GDP (gross domestic product) per capita. Despite these socioeconomic insights, comprehensive research on the spatiotemporal epidemiological characteristics and hospitalisation burden of MMD within the province remains scarce.

Therefore, the purpose of this study was to investigate the spatiotemporal pattern of MMD incidence and hospitalisation burden in Jiangxi Province by using hospitalisation records and geospatial analysis techniques. We conducted a comprehensive analysis of the spatial and temporal characteristics of the epidemiology of MMD and its burden of hospitalisation in the region, aiming to provide a scientific basis for better prevention and treatment of the disease, which is of great significance for formulating effective public health policies and mitigating the burden of MMD. Additionally, the results of this research may offer fresh insights for future epidemiological and etiological investigations of MMD in priority areas.

Method

Standard protocol approvals

This study was conducted in accordance with the Ethical Review Measures for Biomedical Research Involving Human Subjects issued by the National Health Commission of the People’s Republic of China. As a retrospective observational study utilising pre-existing, deidentified administrative hospital discharge data, this research was granted a waiver of individual informed consent by the Review Committee of the Second Affiliated Hospital of Nanchang University (Approval No.: (2021) Yiyan Lun Shen No. (39)).

Data source

We used the Diagnosis Related Groups management system from the Information Centre of the Health Commission of Jiangxi Province to conduct a retrospective observational study on MMD patients admitted to hospitals in Jiangxi Province. Because primary-level hospitals in China typically lack the diagnostic capability for MMD, our data were derived from inpatient records uploaded by 330 secondary and tertiary hospitals in 11 prefecture level cities to the Information Centre of Jiangxi Provincial Health Commission, from which MMD patient records from 210 hospitals were extracted (some specialised hospitals do not treat patients with MMD). Each hospital employs a quality assurance officer who verifies the accuracy of uploaded information to ensure diagnostic confirmation (online supplemental table 1). Population natural changes, population age distribution and consumption level data for Jiangxi Province from 2015 to 2021 were sourced from the Statistical Bureau of Jiangxi Province.

Data extraction

We extracted hospitalisation records uploaded from January 2015 to July 2022 by 210 secondary and tertiary comprehensive hospitals throughout the province and identified patients with a primary diagnosis of MMD, coded as I67.5 according to the International Classification of Diseases, Tenth Revision (ICD-10). The Information Centre standardised, validated and coded the data to ensure completeness and accuracy. Subsequently, two neurologists independently assessed the diagnosis of each potential patient to confirm MMD cases. Patients were excluded if they were diagnosed with diseases associated with moyamoya syndrome (MMS), including atherosclerosis, autoimmune diseases, meningitis, brain tumours, Fabry’s disease, Down syndrome, head trauma or head irradiation. The hospitalisation information included patients’ demographic characteristics, number of hospitalisations, hospitalisation costs, length of stay, comorbidities and complications. The total hospitalisation expenses comprised comprehensive medical service fees, diagnostic fees, treatment fees, medication expenses, consumables fees and other fees. MMD and MMS are distinct clinical entities: MMD refers to idiopathic progressive stenosis of the intracranial internal carotid arteries, whereas MMS occurs secondary to known underlying conditions. To ensure the homogeneity of our study population and accurately reflect the epidemiological features of primary MMD, patients with MMS-associated conditions were excluded in accordance with established diagnostic criteria.

Data screening

Based on the citizen ID number, a unique and invariant legal number in China, we determined the ‘same person’ variable to exclude double counting at subsequent visits. In addition, our study included 18 693 patients after excluding those with missing identity information or non-residents of Jiangxi Province. Patients diagnosed for the first time (new-onset patients) were screened by hospitalisation information. At the same time, the patient’s place of residence was identified according to the citizen ID number, and the lost or unidentifiable patient information was excluded (online supplemental figure 2).

Statistical analysis

Hospitalisation data were collated using Excel 2019 software (Microsoft, USA). In this research, t-test was applied for continuous and normally distributed variables; otherwise, Mann-Whitney U test was employed, with p<0.05 considered statistically significant. In this hospital-based study, annual incidence was defined as the number of newly hospitalised patients per 100 000 population per year. This metric represents a hospitalisation-based detection rate rather than true population incidence, as it does not capture patients managed exclusively in outpatient settings. Several prior studies have similarly used hospitalisation records as a proxy for annual patient counts, an approach that has been empirically justified.^6,8 Age- and sex-adjusted incidence were standardised for each region based on 2021 Jiangxi census data. The annual hospitalisation costs were standardised using the Consumer Price Index. The Mann-Kendall trend test is suitable for analysing time series data with continuous increasing or decreasing trends. It is a non-parametric test, and the p value<0.05 indicates a significant trend in this set of time series data. To achieve more accurate predictions, the fitting curves used in this study were all fitted with cubic functions. R² values range from 0 to 1, with a higher R² indicating a better fit and higher reliability of the trend line. We compared linear, quadratic and cubic fitting models; the cubic function demonstrated the best goodness-of-fit, as indicated by the highest R² values (online supplemental figure 3). Besides, the visual network diagram of comorbidities is constructed by Gephi-0.10 software.

Spatial cluster analysis

Spatiotemporal cluster analysis was selected as the primary analytical approach for this study because our objective was to characterise the geographic distribution and temporal evolution of MMD incidence across spatial units. We identified patient locations based on unique identification numbers and excluded 877 patients with unidentifiable patient locations. The remaining patients were visualised using ArcGIS10.8 (Esri, USA) software, and Moran’s I, Z-score and p value were calculated. We adopted geographic information data for each county as analysed elements and their respective incidence rates as attribute values. Spatial cluster analysis included global and local autocorrelation analysis, hotspot analysis and grouping analysis.⁹ The global autocorrelation analysis used Moran’s I index. When Moran’s I index is 0, it indicates no aggregation. When Moran’s I index is greater than 0, positive spatial autocorrelation exists; otherwise, negative spatial autocorrelation exists. Moran’s I statistical test p value was obtained using Monte Carlo simulation.¹⁰ Local spatial autocorrelation analysis was employed to calculate the correlation between each spatial region and its adjacent regions for a specific attribute, resulting in four categories: high-high clustering areas, high-low scattering areas, low-high scattering areas and low-low clustering areas, or by identifying clustering and outlier patterns through Moran scatterplots. Hotspot analysis calculated the Getis-Ord Gi* statistic for each element, which represents the ‘centre’ of the distribution area and is an indicator of local clustering. This method identified hotspots by comparing local autocorrelation estimates with the overall average level.¹¹ Grouping analysis is based on the attributes of the elements and classifying the regions into different groups, each group possessing a certain degree of similarity and assigned a different colour.

Space-time scan statistics

The SaTScan 9.6 (Martin Kuffdorf, USA) software was applied for space-time scan analysis, which is a widely used method for detecting spatiotemporal clustering of events. It calculates the log-likelihood ratio (LLR) and p value based on the actual and theoretical incidence numbers to identify regional relative risks.¹² When the LLR of the scanning window has statistical significance, it can identify the most likely anomalous space-time event set with clustering characteristics.¹³ In this research, the Space-Time Permutation probabilistic model was initially selected to analyse spatiotemporal rearrangement scan statistics, with dynamically varying cylindrical scan windows for different times and regions, to detect spatiotemporal aggregation in the number and prevalence of MMD.¹⁴ A further spatial aggregation analysis of the number of MMD patients was then carried out using the Purely spatial Poisson model.

Results

The overall incidence and prevalence of moyamoya disease

The average annual incidence from 2015 to 2022 was 3.36 per 100 000 inhabitants, and the average annual prevalence was 5.65 per 100 000. The incidence over this period is increasing (Mann-Kendall trend analysis (p<0.001)), with the highest prevalence of 10.07 per 100 000 inhabitants in 2021, accompanied by the highest incidence rate of 5.47 per 100 000 inhabitants (online supplemental table 2). Based on the cubic fitting curve, the projected incidence for 2022 was 5.30 per 100 000 inhabitants (R²=0.9587) (online supplemental figure 3A). Among the major cities in the province, Nanchang showed the highest incidence and prevalence of 8.01 and 13.01 per 100 000 inhabitants, respectively, followed by Fuzhou with incidence and prevalence of 7.38 and 12.37 per 100 000 inhabitants (online supplemental table 3). Furthermore, we separately explored the annual incidence of each region and observed that the high-incidence cities with an incidence exceeding 4.00 per 100 000 inhabitants in 2022 were Nanchang (11.34), Yichun (6.97), Fuzhou (10.01), Jiujiang (4.17) and Shangrao (4.58).

Epidemiological characteristics of the population

The findings per year for males and females and their average incidence and prevalence are presented in online supplemental figure 3B. Overall, female incidence and prevalence are high compared with male, with the gap most pronounced in 2021 when both metrics reached their peak. The cubic fitting curve projected a female incidence of 5.30 per 100 000 inhabitants for 2022 (R²=0.9581). The male and female incidence by age group for the period 2015–July 2022 is illustrated in online supplemental figure 3C, with the peak incidence of each age group between 51 and 55 years. The incidence for males was 2.29 per 100 000 inhabitants, representing 46.11%, and for females 2.67 per 100 000 inhabitants (online supplemental table 4). In the same age group, prevalence peaked at 3.70 and 4.72 per 100 000 inhabitants for males and females, respectively. Notably, the 6–10 age group exhibited a marked increase in both incidence and prevalence, being three to four times higher than in the previous age group (online supplemental table 5).

Spatially prevalent features

Visualisation of the incident cases by region is displayed in online supplemental figure 4, with a global Moran’s I of 0.39 (Z-score=6.76), indicating a positive spatial autocorrelation and clustered distribution, with less than a 1% probability that this clustering pattern arose by chance (p<0.01). Further local spatial autocorrelation analysis of the spatial distribution of the number of incident cases explored clustering and outliers, hot and cold spot distribution and grouping. The clustering and outlier values indicated high-high clustering in Nanchang, Nanchang Municipal District, Fengcheng, Jinxian and Yugan and generally low-high clustering in the vicinity of these major cities, such as Duchang, Gaoan, Chongren, Yihuang and Yingtan Municipal district. The low-low clusters are mainly found in Yichun Municipal District and Pingxiang Municipal District, while Ganzhou Municipal District forms a high-low cluster with its neighbouring cities. The hot spot analysis revealed that Nanchang and Fuzhou and some of their surrounding cities are hotspots, suggesting that this area is a typical disease cluster. The cluster analysis divided all the districts into four groups, and the number of patients in these four groups exhibited some similarities.

Analogously, this research analysed the overall incidence of MMD in each region. The average incidence in each region for July 2015–2022 is displayed in figure 1A, with the North Central region having a significantly higher incidence than the South West. With a spatial global autocorrelation Moran’s I=0.51 and a Z-score=8.14, the likelihood of randomly generating this clustering pattern is less than 1%, and it is more likely to be clustered and spatially positively correlated than the number of patients. The areas of high and high clustering of incidence were more clearly distributed in Nanchang and Fuzhou and their surrounding cities, but the high and high clustering of the number of patients was more widely distributed than the high and high clustering of the number of patients, which also included several cities south of Fuzhou, including Chongren, Lean, Yihuang and Nanchang. compared with the clustering of patient numbers, the low-incidence and low-incidence clusters are more widely distributed in the western and southern parts of the city (figure 1B). The results for the cold hotspots are more specific, with the highly clustered areas being identified as hotspots (99% confidence) and the five surrounding cities such as Xingan and Nancheng being identified as subhotspots (confidence>90%). In contrast, Ganzhou and Yichun were identified as cold spots (95% confidence) and Luxi and Anyuan as secondary cold spots (90% confidence) (figure 1C). As shown in figure 1D, all study regions were divided into four groups based on their similarity in incidence and spatial correlation.

Figure 1. Spatial correlation analysis of the average incidence of MMD from 2015 to 2022. (A) The average incidence in counties and cities. (B) Clustering and outlier analysis. (C) The hot spot analysis. (D) Group analysis, each colour is assigned by the algorithm to a different group based on its incidence.

Spatiotemporal aggregation analysis

A retrospective space-time scan analysis of the number and incidence of MMD in Jiangxi Province from 2015 to July 2022 was conducted on the software SaTScan v9.6. The outcomes are illustrated in online supplemental figure 5, where there is a clear temporal and spatial clustering of MMD incidence. The results of the retrospective spatiotemporal rearrangement of the number of patients with MMD are presented in online supplemental figure 5A. The most likely cluster was a 50.63 km radius area consisting of Jinxian, Nanchang, Fuzhou Municipal district and Nanchang Municipal district, and the period for the new-onset of MMD was from 1 January 2015 to 31 December 2017. Secondary Cluster 1 comprised a 221.14 km radius area centred on the Ganzhou Municipal District, with a clustering period of 1 January 2020 to 31 December 2021, containing a total of 1648 patients of MMD, with an observed value of 1.23 (p value<0.001). In particular, the Jingdezhen Municipal district from 1 January 2015 to 31 December 2016 was another area of spatiotemporal clustering for MMD (the Secondary Cluster 2), with an observed value of up to 2.66 over the upper expected value (p value<0.001). Similarly, the results of the spatiotemporal rearrangement scan analysis for incidence are summarised in table 1 and visualised in online supplemental figure 5B), with the first three clusters remaining largely consistent with the spatiotemporal clustering of disease counts. However, the urban areas centred on Tonggu, Ruichang and Dean were also detected as subclusters, with the Secondary Cluster 4 of Ruichang and Lushan having the highest mean incidence of 8.35 in 2021 (p value<0.001).

Table 1. The results of the spatiotemporal rearrangement scan analysis for prevalence.

Cluster	Time	Centre	Radius	Average prevalence	Observed\expected	P value
Most likely Cluster	2015–2017	Jinxian, Nanchang, etc	50.63 km	8.03	1.59	<0.001
Secondary Cluster 1	2020–2021	Ganzhou, Shangyou, etc	221.14 km	5.69	1.20	<0.001
Secondary Cluster 2	2015–2016	Jingdezhen	0 km	4.85	3.56	<0.001
Secondary Cluster 3	2015–2016	Tonggu, Yifeng, etc	63.27 km	1.41	1.95	<0.001
Secondary Cluster 4	2021	Ruichang, Lushan	29.93 km	8.35	1.64	<0.001
Secondary Cluster 5	2022	Dean, Gongqingcheng, Yongxiu	33.46 km	6.83	1.47	<0.001

Nevertheless, purely spatial scan analyses are also often used to investigate areas of disease clustering, leaving aside the effect of time. A Poisson model pure spatial scan analysis of the cumulative number of patients from 2015 to July 2022 revealed that a total of nine clusters were detected (table 2). Interestingly, the Secondary Cluster 1 is only part of the first cluster, which includes Jinxian, Nanchang and Fuzhou Municipal district, covering a radius of 41.86 km, but contains 5109 patients, with an observed value of 3.91. It covers a radius of 41.86 km; however, it contains 5109 patients, with an observed value of up to 3.64 compared with the expected value and has the highest relative risk value of 4.70.

Table 2. Results of pure spatial scanning analysis of Poisson’s model for the cumulative number of patients in different regions during 2015–2022 July.

Cluster	Area	Radius	Exposed population	Patients	Observed/expected	RR	LLR	P value
Most likely cluster	Jinxian, Nanchang, Fuzhou Municipal district, Nanchang Municipal district, Fengcheng, Yugan, Jinxi, Chongren	70.08 km	79 606 648	9352	2.38	3.91	3929.60	<0.001
Secondary Cluster 1	Jinxian, Nanchang, Fuzhou Municipal district	41.86 km	28 472 184	5109	3.64	4.70	3350.34	<0.001
Secondary Cluster 2	Zhangshu, Fengcheng, Xingan	35.70 km	14 698 936	1606	2.22	2.34	420.17	<0.001
Secondary Cluster 3	Wannian, Leping, Yugan	38.01 km	15 613 872	1221	1.59	1.63	118.26	<0.001
Secondary Cluster 4	Nanchang Municipal district	0 km	31 437 280	2042	1.32	1.36	78.69	<0.001
Secondary Cluster 5	Anyi, Jingan, Fengxin, Yongxiu	31.48 km	7 674 984	585	1.55	1.57	49.58	<0.001
Secondary Cluster 6	Jiujiang Municipal district	0 km	9 314 144	655	1.43	1.44	38.01	<0.001
Secondary Cluster 7	Guangchang, Nanfeng, Ningdu, Shicheng, Lichuan, Yihuang, Lean, Nancheng, Yongfeng, Chongren	107.81 km	25 223 880	1534	1.23	1.26	34.23	<0.001
Secondary Cluster 8	Yifeng	0 km	2 023 792	165	1.65	1.66	17.92	<0.001

Analysis of clinical complications and comorbidities

The clinical comorbidities and associated complications identified and classified according to ICD-10 are described in table 3 (online supplemental figure 6). 11 140 patients with new-onset MMD had primary clinical complications of haemorrhage, infarction, epilepsy, TIA and others in 41.4%, 32.1%, 2.7%, 4.8% and 19.1% of patients, respectively. Infarction is more common in children, with 29.6% of patients. In contrast, the likelihood of first haemorrhage was greatest in adults, especially in female patients (41.6% in adults and 43.9% in women). In terms of comorbidities, the incidence of risk factors for vascular diseases such as hypertension, diabetes mellitus and hyperlipidaemia was 36.4%, 11.1% and 6.2%, respectively. Additionally, hyperthyroidism or thyroid nodules were seen in approximately 1.5% of patients. Notably, infarction patients had the highest rates of comorbidities, all exceeding 40%.

Table 3. Clinical complications and comorbidities of patients with MMD, 2015–2022.

	Total	Haemorrhage	Infarction	Epilepsy	TIA (TIA: Transient Ischemic Attack)	Others
Demographics
Number, n (%)	11 140 (100)	4609 (41.4)	3573 (32.1)	300 (2.7)	532 (4.8)	2126 (19.1)
Age mean±SD, years	54.127±11.57	51.076±9.98	56.091±11.97	47.793±14.17	55.259±11.67	52.811±12.38
Children, n (%)	108 (1.0)	16 (14.8)	32 (29.6)	15 (13.9)	3 (2.8)	42 (38.9)
Adults, n (%)	11 032 (99.0)	4593 (41.6)	3541 (32.1)	285 (2.6)	529 (4.8)	2084 (18.9)
Female, n (%)	5771 (51.8)	2531 (43.9)	1629 (28.2)	150 (2.6)	281 (4.9)	1180 (20.4)
Comorbidities, n (%)
Hypertension, n (%)	4058 (36.4)	1346 (33.1)	1730 (42.6)	94 (2.3)	251 (6.2)	637 (15.7)
Diabetes, n (%)	1232 (11.1)	190 (15.4)	715 (58.0)	29 (2.4)	64 (5.2)	234 (19.0)
Hyperlipidaemia, n (%)	691 (6.2)	121 (17.5)	333 (48.2)	21 (3.0)	68 (9.8)	148 (21.4)
Hyperthyroidism or thyroid nodules, n (%)	165 (1.5)	29 (17.6)	73 (44.2)	4 (2.4)	10 (6.1)	49 (29.7)

Hospitalisation burden analysis

The average annual length of stay for the 11 127 new-onset patients was 14.073 days, and there was a clear trend towards a year-on-year decrease (table 4). The mean annual hospitalisation cost was CNY 29,409.50, with an overall upward trend observed over the study period, with the highest average annual cost in 2021, expressed as median (IQR): CNY 7,717.03–40,758.168. In terms of payment methods, Urban Employees’ Basic Medical Insurance (UEBMI), Urban Residents’ Basic Medical Insurance (URBMI) and New Rural Cooperative Medical Insurance (NRCMI) account for the vast majority, with URBMI in particular accounting for a high annual average of 28.63%. As the latest initiative by the Chinese government to reform the healthcare system, these three major health insurance schemes cover over 67% of all patients’ payments. Nevertheless, the number of patients who pay entirely out of pocket is still in the minority, with an average annual percentage of 15.59%, which has been increasing especially in the last 4 years. The rest of the patients who have to rely on poverty assistance, commercial insurance or other social insurance payment methods account for only a small percentage.

Table 4. The burden of hospitalisation of MMD patients during 2015–2022.

Year	Total	2015	2016	2017	2018	2019	2020	2021	2022
Number, n (%)	11 127	515	681	914	1419	1956	1791	2470	1381
Hospitalisation stay, mean	14.073	15.909	14.736	15.974	14.509	13.922	14.2	13.424	12.569
Hospitalisation cost, mean	29 409.498	23 773.486	22 211.415	26 652.683	25 981.454	30 005.751	30 907.884	33 424.126	30 439.59
IQR	(6978.695–34 888.23)	(5617.185–28 958.055)	(5768.78–26 786.1)	(6523.955–37 148.73)	(6487.59–31 887.865)	(7339.038–36 759.808)	(6892.21–34 738.75)	(7717.03–40 758.168)	(7562.18–35 384.63)
Payment methods
UEBMI	2196 (19.84%)	73 (14.17%)	99 (14.54%)	169 (18.51%)	315 (22.21%)	447 (22.93%)	375 (21.07%)	605 (24.70%)	113 (8.29%)
URBMI	3169 (28.63%)	45 (8.74%)	109 (16.01%)	205 (22.45%)	455 (32.09%)	671 (34.43%)	601 (33.76%)	641 (26.17%)	442 (32.43%)
NRCMI	2101 (18.98%)	250 (48.54%)	281 (41.26%)	272 (29.79%)	280 (19.75%)	267 (13.70%)	251 (14.10%)	303 (12.37%)	197 (14.45%)
Poverty relief	259 (2.34%)	0 (0.00%)	0 (0.00%)	6 (0.66%)	41 (2.89%)	53 (2.72%)	51 (2.87%)	83 (3.39%)	25 (1.83%)
χ	262 (2.37%)	3 (0.58%)	43 (6.31%)	31 (3.40%)	45 (3.17%)	59 (3.03%)	56 (3.15%)	20 (0.82%)	5 (0.37%)
All public expense	163 (1.47%)	7 (1.36%)	0 (0.00%)	21 (2.30%)	20 (1.41%)	19 (0.97%)	38 (2.13%)	44 (1.80%)	14 (1.03%)
Fully self-funded	1726 (15.59%)	41 (7.96%)	34 (4.99%)	98 (10.73%)	218 (15.37%)	366 (18.78%)	258 (14.49%)	468 (19.11%)	243 (17.83%)
Other social insurance	62 (0.56%)	40 (7.77%)	1 (0.15%)	3 (0.33%)	2 (0.14%)	1 (0.05%)	5 (0.28%)	4 (0.16%)	6 (0.44%)
Other	1130 (10.21%)	56 (10.87%)	114 (16.74%)	108 (11.83%)	42 (2.96%)	66 (3.39%)	145 (8.15%)	281 (11.47%)	318 (23.33%)

CHI, Commercial Health Insurance; IQR, interquartile range; NRCMI, New Rural Cooperative Medical Insurance; UEBMI, Urban Employee Basic Medical Insurance; URBMI, Urban Resident Basic Medical Insurance.

Discussion

Principal findings

This study represents the most comprehensive analysis of MMD epidemiology in Jiangxi Province, the Chinese province with the highest reported MMD incidence. Using hospitalisation records from 210 secondary and tertiary hospitals spanning 2015 to mid-2022, we identified 18 693 hospitalisation records, including 11 140 new-onset patients. The average annual hospitalisation-based incidence was 3.36 per 100 000 inhabitants, with a peak annual incidence of 5.47 per 100 000 in 2021. This figure is nearly double the rate of 2.81 per 100 000 previously reported in a nationwide hospital-based study,⁷ suggesting a continued upward trend in MMD detection in this high-incidence region. Spatial analysis revealed a significant clustered distribution pattern (Moran’s I=0.51, Z-score=8.14, p<0.001), with Nanchang and Fuzhou identified as persistent hotspots.

Comparison with previous studies

Our findings extend and update previous epidemiological data on MMD in China. The incidence rate observed in Jiangxi Province substantially exceeds rates reported in other East Asian populations, including Japan (0.54 per 100 000 in 2003)⁴ and Korea (0.35 per 100 000 in 2004),¹⁵ as well as the national Chinese average of 1.14 per 100 000 reported for 2016–2018.⁷ Several factors may contribute to this disparity. First, the increasing availability and resolution of non-invasive neuroimaging techniques, particularly magnetic resonance angiography, has improved MMD detection rates globally.¹ Second, the publication of the first Chinese clinical guidelines for MMD diagnosis and treatment in 2017⁸ likely enhanced diagnostic awareness among clinicians. Third, Jiangxi Province may harbour region-specific genetic or environmental risk factors that warrant further investigation. It is important to note that our hospital-based study design, which captures only hospitalised patients, may underestimate the true population incidence by excluding cases managed exclusively in outpatient settings. Conversely, the concentration of tertiary medical centres in Nanchang and Fuzhou may lead to referral-related overestimation in these cities.

Pidemiological characteristics

Our analysis revealed a modest gender disparity, with female patients exhibiting higher overall incidence and prevalence than males (2.67 vs 2.29 per 100 000). The female-to-male incidence ratio of approximately 1.17:1 is notably lower than ratios reported in other Asian populations (1.8–2.2:1)¹⁶ and Western countries (1.8–4.25:1),^17,18 but consistent with findings from Taiwan (1.4:1)¹⁹ and Nanjing (1.1:1).⁵ These variations in sex ratios across populations suggest that genetic factors play a significant role in MMD pathogenesis, as previously proposed.²⁰ The peak incidence in adults occurred between ages 51 and 55 years, corroborating earlier findings.²¹ A secondary, less pronounced peak was observed in children aged 11–15 years. The relatively older age at peak incidence compared with some prior reports may reflect cohort effects, improved diagnostic recognition in older adults or changes in the demographic profile of affected individuals over time.

Spatiotemporal distribution and clustering

In terms of regional distribution, the incidence and prevalence of MMD in Nanchang and Fuzhou are significantly higher than in other cities, with a pronounced positive regional spatial correlation and clustering effect, particularly around Nanchang (online supplemental table 2). Local spatial autocorrelation analysis suggests that MMD control and prevention efforts should be intensified in persistent high-incidence clusters and hotspots, including Nanchang, Yugan, Fengcheng, Chongren, Lean and Yihuang. For cities with varying incidence levels, such as Yingtan, Gaoan and Duchang, enhanced MMD screening is recommended. The cluster analysis results can guide policymakers and researchers in assessing the disaggregated prevalence of MMD across cities, identifying shared disease characteristics, and informing targeted screening and prevention strategies.

Currently, scanning statistics are extensively used for detecting temporal and spatial clustering of events, facilitating early warning systems to optimise data mining and aggregate detection.²² The primary distinction between the two models lies in the Poisson model’s requirement for population data for the study area, while the Space-Time Permutation model obviates this need by employing a rearrangement algorithm.^23,24 This study employs both models to analyse the distribution of MMD patients in terms of numbers and prevalence. For areas with high prevalence, a spatiotemporal rearrangement scan identified a disease outbreak in Jinxian, occurring between 2015 and 2017 within a 50.63 km radius, with an average prevalence of 8.03. A spatial scan using the Poisson model pinpointed a more refined area of relative risk within a 41.86 km radius of Jinxian, where the relative risk factor peaked at 4.70. Consequently, it is recommended that areas exhibiting high spatial and temporal clustering be subject to intensified surveillance to prevent a resurgence of the epidemic.

Clinical complications and comorbidities

The clinical profile of MMD patients in Jiangxi Province revealed notable patterns. Haemorrhage was the most common presenting complication (41.4%), followed by infarction (32.1%) and transient ischaemic attack (4.8%). The predominance of haemorrhagic presentations contrasts with some Western cohorts where ischaemic events predominate but is consistent with reports from other Chinese populations.⁶ Among comorbidities, hypertension (36.4%) was the most prevalent, followed by diabetes mellitus (11.1%) and hyperlipidaemia (6.2%). The diabetes prevalence exceeded rates reported in prior large-scale Chinese MMD studies, which may reflect the rising background prevalence of diabetes in the general Chinese population or improved screening practices. Prior research has identified preoperative hypertension as a significant risk factor for adverse surgical outcomes in MMD patients,^25,26 underscoring the importance of vigilant blood pressure management. Additionally, 1.5% of patients had comorbid hyperthyroidism or thyroid nodules, consistent with emerging evidence linking thyroid dysfunction to MMD.²⁷ These findings highlight the need for comprehensive metabolic screening in MMD patients.

Hospitalisation burden

The mean annual hospitalisation cost was RMB 29,409, with an upward trend over the study period. Despite rising costs, more than 67% of patients were covered by basic medical insurance (UEBMI, URBMI or NRCMI), reflecting the expanding reach of China’s healthcare insurance system. However, the proportion of fully self-funded patients showed an increasing trend in recent years (from 8.0% in 2015 to 17.8% in 2022), which may indicate gaps in insurance coverage or an influx of patients from outside the province who are not covered by Jiangxi’s insurance programmes. The mean length of hospital stay showed a decreasing trend (from 15.9 days in 2015 to 12.6 days in 2022), potentially reflecting improvements in surgical techniques, perioperative care and hospital efficiency. These hospitalisation burden data provide important benchmarks for healthcare resource planning in high-incidence regions.

Potential etiological implications

The RNF213 variant plays a pivotal role in the development of MMD. Prior research has indicated that provinces in China with a higher prevalence of the RNF213 p.R4810K mutation are predominantly located in the eastern and northeastern coastal regions. Jiangxi, situated in the East China region, exhibits a significant carrier rate of 29.5%, surpassing all other areas except Northeast China (32.0%).²⁸ Our study suggests that, in addition to genetic factors, there may be other unique elements significantly influencing the incidence of MMD in Jiangxi Province. Interestingly, an examination of topographical features in Jiangxi Province reveals a higher incidence of MMD in cities surrounding the Poyang Lake Basin, with a north-high-and-south-low pattern, offering insights for future epidemiological and etiological research (online supplemental figure 1). The observed disparities in MMD epidemiology in Jiangxi, compared with other regions, may be attributed to various environmental factors and medical resource disparities. These geographic and ecological commonalities present novel perspectives for subsequent etiological inquiries.

The treatment and management of moyamoya disease

Recent advances in the treatment and management of MMD warrant discussion in the context of our epidemiological findings. Surgical revascularisation, including direct bypass (superficial temporal artery-to-middle cerebral artery anastomosis), indirect bypass (encephalo-duro-arterio-synangiosis) and combined bypass procedures, remains the cornerstone of MMD treatment.²⁹ Growing evidence suggests that early surgical intervention in symptomatic patients can significantly reduce the risk of subsequent ischaemic and haemorrhagic stroke events.³⁰ For regions identified as high-incidence clusters in our study, such as Nanchang and Fuzhou and their surrounding areas, targeted screening programmes and the establishment of specialised MMD treatment centres could facilitate earlier diagnosis and timely surgical intervention. Furthermore, the identification of hypertension as the most prevalent comorbidity (36.4%) among MMD patients in our cohort underscores the critical importance of perioperative blood pressure management, which has been identified as a significant risk factor for surgical outcomes in MMD patients. The emerging role of genetic screening, particularly for the RNF213 p.R4810K variant, which has been shown to have a high carrier rate (29.5%) in Jiangxi Province, may offer opportunities for early identification of at-risk individuals in high-incidence areas. These clinical insights, combined with our epidemiological findings, provide a framework for developing targeted public health strategies to reduce the burden of MMD in high-risk regions.

Conclusions

In summary, this study uses hospital discharge data and geospatial analysis techniques to explore, for the first time, the spatial and temporal patterns of MMD and the burden of hospitalisation in Jiangxi Province with greater detail and comprehensive accuracy. Nevertheless, the use of hospital admission data as a proxy for incidence may introduce several potential biases: (1) underestimation of true incidence, as patients managed solely in outpatient settings are not captured; (2) geographic access bias, whereby regions with better healthcare infrastructure may report higher rates; (3) socioeconomic bias, as patients with greater financial resources may be more likely to seek hospital care and (4) referral bias, as some patients, particularly children, may travel to other provinces for specialised medical care. These have led to an underestimation of incidence and prevalence. Our research group will obtain the medical insurance data covering the whole province from the government in the future to further improve and confirm our research results. To sum up, the results of this study are still important for understanding the epidemiological characteristics of the provinces with the highest prevalence of MMD in China and for developing effective public health policies to reduce the burden of MMD.

Data availability statement

Data are available upon reasonable request.