Association of initial diagnostic setting and referral delay with mortality in systemic sclerosis: a nationwide database cohort study.

Introduction

Systemic sclerosis (SSc) is a rare disease characterized by skin tightness. Disease progression has been described in three phases: edematous, indurative, and atrophic [ 1 ]. The edematous phase is the early phase of the disease, and the initial presentations could be puffy hands, sclerodactyly, and/or Raynaud’s phenomenon [ 1 ]. The clinical manifestations in this phase make it difficult to make the diagnosis because classical skin tightness cannot be detected in the early phase of the disease. Skin tightness is the classical presentation in the indurative phase of the disease, and its extension is classified into two major subsets: limited cutaneous systemic sclerosis (lcSSc) and diffuse cutaneous systemic sclerosis (dcSSc) [ 2 ]. Internal organ involvement could be detected early in the course of the disease, particularly in dcSSc [ 3 ], and patients with internal organ involvement trended to be referred to secondary or tertiary centers [ 4 , 5 ]. Since internal organ involvement is associated with significant morbidity and mortality [ 3 , 6 ], these patients should be closely followed up to monitor internal organ involvement and early treatment initiation. Emerging evidence suggests that the timing of diagnosis and initiation of disease-modifying therapies, particularly for interstitial lung disease (ILD) and pulmonary arterial hypertension (PAH), critically impacts long-term outcomes [ 7 ].

Previously, Foocharoen et al. [ 5 ] reported that patients with SSc who were referred early had better prognoses than those referred later. However, the findings of this study cannot be generalized to the entire country. Thailand’s health policy has developed Universal Coverage for all Thai citizens who do not have other health insurance or access to other health services. Patients under the Universal Coverage scheme must initially use health services at primary care hospitals (Subdistrict Health Promotion Hospitals), secondary care hospitals (Community Hospitals), or registered affiliated hospitals (mostly secondary hospitals). If specialist care is required, patients are referred to tertiary care centers (General Hospitals and Regional Hospitals) or to university hospitals [ 8 ]. Understanding factors that influence diagnostic pathways and referral patterns is essential for optimizing outcomes in this complex disease.

Currently, data are lacking on the proportion of patients with SSc who were initially diagnosed at secondary hospitals prior to referral to tertiary care centers, as well as their referral duration and subsequent outcomes. Because this study was based on an administrative database, the onset of symptoms and true diagnostic delay could not be directly determined. Therefore, this study focused on referral pathways and timing within the healthcare system. Understanding these referral patterns is important for optimizing specialist access and improving outcomes in patients with SSc. We aimed to evaluate the association of initial diagnostic setting and referral duration with mortality among patients with SSc in Thailand.

Methods

This was a cohort study conducted by using the National Health Security Office (NHSO) database which captures approximately 48 million beneficiaries, representing roughly 75% of Thailand’s population, providing robust population-level data for rare disease epidemiology [ 9 ]. We included patients aged ≥ 18 years with SSc (ICD-10 code M34) recorded at secondary hospitals, tertiary hospitals, or university hospitals from October 1, 2016, to December 31, 2023. Patients could enter the cohort through direct diagnosis at tertiary/university hospitals or through referral from secondary hospitals within the referral system. Follow-up duration varied across patients depending on the date of cohort entry and continued until death or censoring on December 31, 2024. We excluded patients with any of the following: overlap with other connective tissue diseases or referral for a diagnosis other than SSc. The study period and data extraction were aligned with the national fiscal year (October 1 to September 30) to ensure consistency with the NHSO reporting cycles.

All data of eligible patients were traced back to the primary referral hospitals. The following data were extracted from tertiary care hospitals or university hospitals: hospital center, sex, date of first referral, primary and secondary diagnoses, coexisting diseases, and outcome status on December 31, 2024 (survival or death). The data extracted from secondary care hospitals included the hospital center, referral date, and primary diagnosis on the referral date. Common coexisting conditions were extracted using International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10), including diabetes mellitus (E11), hypertension (I10), lung cancer (C34), acute myocardial infarction (I21), chronic kidney disease (N18), respiratory disorders in other diffuse connective tissue diseases (J99.1), other interstitial pulmonary diseases (J84), and primary pulmonary hypertension (I27.0). Data quality is maintained through mandatory reporting requirements and regular audits. ICD-10 coding is performed by trained medical records staff following international classification standards.

Operational definitions

SSc cases were identified using ICD-10 M34 codes recorded in the NHSO database. Patients could have SSc diagnoses recorded at secondary hospitals, tertiary hospitals, or both within the referral pathway. For patients referred from secondary hospitals, prior SSc diagnoses at the secondary level were identified from earlier ICD-10 M34 records before tertiary referral. Cohort entry and follow-up were defined based on the first recorded diagnosis at a tertiary or university hospital when available. Coexisting diseases were also identified from ICD-10 codes recorded within the same administrative database before or at cohort entry. Referred patients were those referred from any hospital. The start date was the date on which patients were first diagnosed with SSc at any tertiary hospital. The end date was defined as the date of death from any cause; otherwise, it was December 31, 2024, for patients who were still alive. To approximate incident SSc cases and reduce survivor bias, a 1-year washout period was applied using available NHSO data from October 1, 2015, to September 30, 2016. Patients with any prior ICD-10 M34 diagnosis recorded during this period were excluded. Therefore, only patients with the first recorded SSc diagnosis from October 1, 2016, onward were included in the cohort. The referral date was the date of referral from secondary care hospitals. Referral duration was specifically defined as the time interval between the first recorded SSc diagnosis at a secondary hospital and the first recorded diagnosis at our tertiary center, rather than the total duration of symptoms. This measure reflects referral timing within the administrative database’s recorded healthcare pathway and should not be interpreted as the true disease onset or total duration. We categorized patients based on their diagnostic pathway. Patients “first diagnosed at a secondary care hospital” before being referred represent the group where initial recognition occurred outside a specialized center, whereas the comparison group consisted of patients whose first recorded SSc diagnosis occurred at a tertiary or university hospital. Time-to-event (death) was defined as the interval from the start date to the end date (death or censoring date).

Statistical analysis

Electronic health records and diagnostic data for all patients diagnosed with SSc were extracted from the NHSO database and systematically analyzed. The clinical characteristics were categorized according to the theory being studied. Each data point was assigned as a dichotomous, polytomous, or continuous variable. The proportion of patients with delayed diagnosis at secondary hospitals prior to referral and the time-to-referral were calculated and compared with those who were appropriately referred and correctly diagnosed. Factors associated with delayed diagnosis (such as age, sex, and coexisting diseases) were explored using the Chi-square test for categorical data and the Student’s t-test or Wilcoxon rank-sum test for continuous data, as appropriate. Patient outcomes (survival or death) were ascertained through electronic linkage with the national death registry and NHSO databases. The mortality rate was reported with 95% confidence intervals (95% CI) and described separately for patients first diagnosed with SSc in secondary and tertiary care hospitals. Survival probabilities were estimated using the Kaplan-Meier method, and differences between groups were compared using the log-rank test. Mortality risk factors (including age, sex, time to referral, and coexisting diseases at baseline) were explored in both groups: those with delayed diagnosis at secondary care hospitals and those correctly diagnosed at tertiary or university hospitals.

Two multivariable Cox regression models were constructed. The first model included the overall SSc cohort and evaluated the association between initial diagnostic setting (secondary versus tertiary/university hospital) and mortality. The second model was restricted to patients initially diagnosed at secondary hospitals and evaluated the association between referral duration and mortality among referred patients.

Cox proportional hazards modeling with Schoenfeld residuals was used to evaluate mortality risk. Parameters that violated the proportional hazards assumption were handled using time-dependent effects or stratified Cox regression, depending on the type of variable. For continuous variables, non-proportional effects were modeled by incorporating interaction terms with log(time). Specifically, age was modeled using a time-dependent interaction term between age and log(time). Sex, hypertension, lung cancer, and chronic kidney disease stage were included as stratification variables to allow for non-proportional baseline hazards. Other covariates (referral status, diabetes mellitus, acute myocardial infarction, and respiratory conditions) were included as fixed effects.

In comparison, stratified Cox regression was employed to accommodate different baseline hazards across groups without imposing the assumption of a constant hazard ratio throughout the follow-up period when the parameters were categorical. To assess the robustness of the primary findings, several additional analyses were performed. A sensitivity analysis excluded patients who died within 30 days of diagnosis at the tertiary hospital to address potential bias from early mortality. Furthermore, to evaluate the consistency of the association between referral delay and mortality, time-to-referral was examined using alternative cutoff values (7, 14, 30, and 60 days) in adjusted Cox models. These analyses allowed us to verify whether the observed effects persisted across different definitions of delay. All regression models were adjusted for relevant covariates, and results are presented as adjusted hazard ratios with 95% confidence intervals and p-values. All statistical tests were two-tailed, and a p-value of < 0.05 was statistically significant. All data analyses were performed using STATA version 16.0 (StataCorp., College Station, TX, USA).

Results

A total of 17,036 patients with SSc were included in the study. Most patients (11,585 cases, 68%) were female. The mean age was 56.3 ± 12.1 years. Hypertension was the most common coexisting disease (40.8%), followed by diabetes mellitus (15.9%), and chronic kidney disease stage 3 to 5 (9.3%). The overall median follow-up duration was 46.8 months, with an interquartile range of 18.6 to 77.5 months. Nearly 75% of the cases were first diagnosed in tertiary care hospitals, and 36.3% did not survive the follow-up period. The demographic data are presented in Table  1 .

Table 1 Patients’ demographic data Demographic N  = 17,036 Sex, n (%)  Male 5,451 (32.0)  Female 11,585 (68.0)  Age (years), mean (SD) 56.3 (12.1) Secondary diagnosis, n (%)  Diabetes Mellitus 2,701 (15.9)  Hypertension 6,946 (40.8)  Lung cancer 306 (1.8)  Acute myocardial infarction 599 (3.5)  Chronic kidney disease stage 3 to 5 1,592 (9.3)  Respiratory disorders in other diffuse connective tissue disorders 221 (1.3)  Other interstitial pulmonary diseases 812 (4.8)  Primary pulmonary hypertension 129 (0.8) Levels of hospital at the first diagnosis, n (%)  Secondary care hospitals 4,404 (25.9)  Tertiary care hospitals 12,632 (74.2) Clinical outcomes, n (%)  Survive 10,875 (63.3)  Non-survive 6,161 (36.2) SD Standard deviation

Patients’ demographic data

SD Standard deviation

A total of 4,404 patients with SSc from 102 tertiary care hospitals were referred from secondary care hospitals (Fig.  1 .). Almost 60% were referred by regional and non-government hospitals. Most patients were female (2,770 cases; 62.9%). The respective mean age and median time to referral were 56.8 ± 10.8 years and 35 days (range 1–2,637). The median follow-up time was 49.2 months (IQR 19.7–79.7 months) among patients first diagnosed at tertiary hospitals and 39.5 months (IQR 15.7–70.6 months) among those first diagnosed at secondary hospitals. Nearly 40% of the patients had coexisting hypertension. Compared to patients who were first diagnosed at tertiary hospitals, patients who were first diagnosed at secondary hospitals were more frequently male, older, and non-survivors, but with a lower frequency of diabetes mellitus and hypertension than in patients who were first diagnosed at tertiary care hospitals (Table  2 ).

Flow of patients

Table 2 Data comparison between the patients who were diagnosed with SSc at secondary care hospitals and tertiary care hospitals Demographic First diagnosis at secondary care hospitals First diagnosis at tertiary care hospitals p -value N  = 4,404 N  = 12,632 Sex, n (%) < 0.001 a  Male 1,634 (37.1) 3,817 (30.2)  Female 2,770 (62.9) 8,815 (69.8) Age (years), mean (SD) 56.8 (10.8) 56.1 (12.5) < 0.001 a Referral duration  Within 7 days 1,370 (31.1) -  8–30 days 753 (17.1) -  31–180 days 1,076 (24.4) -  > 180 days 1,205 (27.4) -  Median (days) (min-max) 35 (1-2637) - Secondary diagnosis  Diabetes Mellitus, n (%) 626 (14.2) 2,075 (16.3) < 0.001 a  Hypertension, n (%) 1,690 (38.4) 5,256 (41.6) < 0.001 a  Lung cancer, n (%) 86 (2.0) 220 (1.7) 0.36  Acute myocardial infarction, n (%) 152 (3.4) 447 (3.5) 0.79  Chronic kidney disease stage 3 to 5, n (%) 388 (8.8) 1,204 (9.5) 0.16  Respiratory disorders in other diffuse connective tissue disorders 60 (1.4) 161 (1.3) 0.66  Other interstitial pulmonary diseases 221 (5.0) 591 (4.7) 0.36  Primary pulmonary hypertension 38 (0.9) 91 (0.7) 0.35 Type of referral hospital, n (%)  General Hospital 1,794 (40.7) -  Regional and non-governmental hospital 2,610 (59.3) - Clinical status < 0.001 a  Survive 2,630 (59.7) 8,245 (65.3)  Non-survive 1,774 (40.3) 4,387 (34.7) a Statistical significance SD Standard deviation

Data comparison between the patients who were diagnosed with SSc at secondary care hospitals and tertiary care hospitals

a Statistical significance

SD Standard deviation

During the follow-up period, a total of 850,897.44 person-months were accumulated. The follow-up duration for patients initially diagnosed at secondary hospitals and tertiary hospitals was 200,117 and 650,780 person-months, respectively. A total of 6,161 deaths were observed, corresponding to a mortality rate of 0.72 per 100 person-months (95% CI: 0.70–0.74). The Kaplan-Meier survival graph is presented in Fig.  2 .

Kaplan-Meier survival graph

The mortality rate of patients first diagnosed in secondary care hospitals (0.89 per 100 person-months, 95%CI 0.85–0.93) was significantly higher than that in tertiary care hospitals (0.67 per 100 person-months, 95%CI 0.65–0.69), ( p  < 0.001). The median survival of patients first diagnosed in secondary care hospitals was 92.2 months, and that of patients first diagnosed in tertiary care hospitals was greater than 100 months. A Kaplan-Meier survival graph comparing patients first diagnosed in secondary and tertiary care hospitals is shown in Fig.  3 .

Kaplan-Meier survival graph categorized by the level of hospital at the first diagnosis

Univariable analysis showed that the mortality risks were older age, diabetes mellitus, hypertension, lung cancer, acute myocardial infarction, chronic kidney disease stage 3 to 5, respiratory disorders in other diffuse connective tissue disorders, other interstitial pulmonary diseases, primary pulmonary hypertension, and first diagnosis at secondary care hospitals. In addition, female sex was a protective factor for mortality (Table  3 ).

Table 3 Mortality risk of overall SSc patients by univariable analysis Variables Crude HR (95%CI) Sex, n (%)  Male 1  Female 0.664 (0.631–0.699) a Every 1-year increase in age 1.041 (1.039–1.044) a Secondary diagnosis, n (%)  Diabetes Mellitus 1.139 (1.066–1.217) a  Hypertension 1.055 (1.003–1.109) a  Lung cancer 1.898 (1.642–2.194) a  Acute myocardial infarction 2.273 (2.050–2.520) a  Chronic kidney disease stage 3 to 5 1.648 (1.531–1.773) a  Respiratory disorders in other diffuse connective tissue disorders 1.250 (1.005–1.554) a  Other interstitial pulmonary diseases 1.235 (1.108–1.378) a  Primary pulmonary hypertension 1.913 (1.530–2.392) a Level of hospital at the first diagnosis, n (%)  Secondary care hospitals 1.275 (1.207–1.347) a  Tertiary care hospitals 1 a Statistical significance

Mortality risk of overall SSc patients by univariable analysis

a Statistical significance

Based on the Cox proportional hazards model, the proportional hazards assumption was violated for age, sex, hypertension, lung cancer, and chronic kidney disease stages 3–5. Therefore, a time-dependent interaction term between age and log(time) was incorporated to account for the non-proportional effect of age over time, whereas the remaining categorical variables were handled using stratified Cox regression to allow for different baseline hazards across strata. In the adjusted analysis, acute myocardial infarction, respiratory disorders in other diffuse connective tissue disorders, other interstitial pulmonary diseases, primary pulmonary hypertension, and initial diagnoses at secondary care hospitals were significantly associated with an increased mortality risk. Primary pulmonary hypertension showed the strongest association with mortality (adjusted hazard ratio (HR) 1.94, 95% CI 1.53–2.46). The results of the Cox regression analysis for overall SSc patients are presented in Table  4 .

Table 4 Mortality risk of overall SSc patients by Cox regression analysis Variables Adjusted HR (95%CI) Every 1-year increase in age 1.053 (1.048–1.060) a Secondary diagnosis, n (%)  Diabetes Mellitus 1.054 (0.981–1.133)  Acute myocardial infarction 1.879 (1.691–2.089) a  Respiratory disorders in other diffuse connective tissue disorders 1.344 (1.062–1.701) a  Other interstitial pulmonary diseases 1.247 (1.112–1.399) a  Primary pulmonary hypertension 1.937 (1.525–2.460) a Level of hospital at the first diagnosis, n (%)  Secondary care hospitals 1.244 (1.177–1.316) a  Tertiary care hospitals 1 a Statistical significance, analyzed using a time-dependent interaction term between age and log(time); the hazard ratio was 0.995 per additional month (a 0.5% decrease per month)

Mortality risk of overall SSc patients by Cox regression analysis

a Statistical significance, analyzed using a time-dependent interaction term between age and log(time); the hazard ratio was 0.995 per additional month (a 0.5% decrease per month)

Among patients initially diagnosed in secondary care hospitals and subsequently referred to tertiary care hospitals, the time to referral significantly influenced the mortality rate. For this subgroup, the follow-up duration was 69,211 person-months for those referred within 7 days, 31,417.6 person-months for 8–30 days, 45,563.7 person-months for 31–180 days, and 45,992 person-months for those referred after more than 180 days. A shorter referral duration was associated with a lower mortality rate than a longer referral duration. The Kaplan-Meier survival curves stratified by time to referral are presented in Fig.  4 . Parameters associated with mortality risk among patients first diagnosed in secondary care hospitals, as determined by univariable analysis, are presented in Table  5 . The Cox proportional hazard model showed assumption violations for age, sex, hypertension, lung cancer, and chronic kidney disease stages 3–5. Therefore, age was analyzed using a time-dependent interaction term with log(time), while the others were analyzed using stratified Cox regression. The stratified Cox regression analysis showed that older age, prolonged referral duration, acute myocardial infarction, and primary pulmonary hypertension were associated with mortality in patients with SSc who were first diagnosed at a secondary care hospital (Table  6 ).

Kaplan-Meier survival graph categorized by the referral duration

Table 5 Mortality risk of SSc patients who were first diagnosed at secondary care hospitals by univariable analysis Variables Crude HR (95%CI) Sex, n (%)  Male 1  Female 0.700 (0.637–0.769)* Every 1-year increase in age 1.041 (1.036–1.045)* Referral duration  Within 7 days 1  8–30 days 1.247 (1.084–1.435) a  31–180 days 1.269 (1.120–1.439) a  > 180 days 1.254 (1.107–1.421) a Secondary diagnosis, n (%)  Diabetes Mellitus 1.063 (0.932–1.211)  Hypertension 0.986 (0.896–1.085)  Lung cancer 1.646 (1.255–2.158) a  Acute myocardial infarction 2.163 (1.781–2.628) a  Chronic kidney disease stage 3 to 5 1.473 (1.275–1.703) a  Respiratory disorders in other diffuse connective tissue disorders 1.181 (0.783–1.783)  Other interstitial pulmonary diseases 1.004 (0.812–1.243)  Primary pulmonary hypertension 1.651 (1.104–2.470) a a Statistical significance

Mortality risk of SSc patients who were first diagnosed at secondary care hospitals by univariable analysis

a Statistical significance

Table 6 Mortality risk of SSc patients who were first diagnosed at secondary care hospitals by Cox regression analysis Variables Adjusted HR (95%CI) Every 1-year increase in age a 1.055 (1.045–1.064) a Referral duration  Within 7 days 1  8–30 days 1.245 (1.081–1.434) a  31–180 days 1.236 (1.089–1.403) a  > 180 days 1.300 (1.141–1.471) a Secondary diagnosis, n (%)  Diabetes Mellitus 1.033 (0.897–1.190)  Acute myocardial infarction 1.895 (1.550–2.315) a  Respiratory disorders in other diffuse connective tissue disorders 1.115 (0.737–1.687)  Other interstitial pulmonary diseases 1.037 (0.829–1.294)  Primary pulmonary hypertension 1.654 (1.069–2.559) a a Statistical significance, analyzed using a time-dependent interaction term between age and log(time); the hazard ratio was 0.995 per additional month (a 0.5% decrease per month)

Mortality risk of SSc patients who were first diagnosed at secondary care hospitals by Cox regression analysis

a Statistical significance, analyzed using a time-dependent interaction term between age and log(time); the hazard ratio was 0.995 per additional month (a 0.5% decrease per month)

When the sensitivity analysis was restricted to patients with a follow-up time greater than 30 days to address potential bias from early mortality, the association between initial diagnosis at secondary hospitals and mortality remained significant (adjusted HR 1.22, 95% CI 1.11 to 1.34, p  < 0.001). Similarly, referral delays exceeding 7 days continued to demonstrate increased mortality risk in this restricted cohort. When alternative referral delay thresholds were examined, delays exceeding 14 days (adjusted HR 1.25, 95% CI 1.13 to 1.38), 30 days (adjusted HR 1.19, 95% CI 1.08 to 1.31), and 60 days (adjusted HR 1.17, 95% CI 1.06 to 1.29) all showed consistent associations with increased mortality, supporting the robustness of the primary findings.

Discussion

In this large nationwide cohort study, we investigated the association between the level of the hospital at the time of initial diagnosis and mortality in patients with SSc. Patients first diagnosed at secondary care hospitals had significantly higher mortality rates and shorter median survival than those initially diagnosed at tertiary care centers. Early recognition and specialized diagnosis of SSc may significantly influence long-term outcomes.

In this nationwide cohort of 17,036 patients with SSc under Thailand’s Universal Coverage scheme, we found that the level of hospital at first SSc diagnosis and the time of referral from secondary to tertiary care were strongly associated with patient survival. Overall mortality was 0.72 per 100 person-months, and patients whose first SSc diagnosis occurred at secondary care hospitals had a significantly higher mortality rate than those first diagnosed at tertiary care centers, even after adjusting for age and major comorbidities. For patients initially diagnosed in secondary care, delayed referral to tertiary centers was an independent predictor of increased mortality. Notably, all referral delays exceeding seven days carried a similar excess risk; this pattern likely reflects baseline illness severity or the necessity for emergency presentation rather than being a sole indicator of referral quality. In addition, acute myocardial infarction, respiratory manifestations, and primary pulmonary hypertension were major determinants of poor outcomes. These findings complement and extend previous Thai data, suggesting that early referral of patients with SSc is associated with a better prognosis. Foocharoen et al. reported that patients referred earlier to specialist centers had more favorable outcomes than those referred later; however, their findings were limited to specific centers and could not be generalized nationwide [ 5 ]. Our study, using the National Health Security Office database and including over 17,000 patients from tertiary care hospitals, confirms at a national level that patients managed within the referral network experience substantial mortality and that both the point at which SSc is first recognized and the speed of referral matter for survival.

Early specialist referral remains a critical challenge in SSc management. Recent data suggest that a definitive diagnosis remains protracted, often spanning over three years from the initial presentation of Raynaud’s phenomenon, as patients frequently navigate multiple healthcare providers before reaching a specialist [ 10 ]. This diagnostic lag is increasingly problematic given that the 2024 British Society for Rheumatology guidelines [ 11 ] and the 2023 the European Alliance of Associations for Rheumatology (EULAR) [ 12 ] update now advocate for earlier, more aggressive interventions—such as immediate nailfold capillaroscopy and upfront combination therapies—to mitigate the high mortality associated with cardiopulmonary involvement.

The observed mortality in this cohort (36.3% non-survivors during follow-up; overall mortality rate, 0.72 per 100 person-months) is consistent with SSc being a high-fatality disease, particularly when internal organ involvement is present. Previous international cohorts have reported 10-year survival rates between roughly 60–80%, with worse outcomes in diffuse cutaneous SSc and in patients with cardiopulmonary or renal involvement [ 6 ]. Although our administrative dataset did not distinguish limited from diffuse cutaneous SSc or detail organ involvement, the strong associations observed for acute myocardial infarction, other interstitial pulmonary diseases, and primary pulmonary hypertension likely reflect the burden of cardiac and pulmonary involvement that is known to drive morbidity and mortality in SSc [ 6 , 13 ].

Mortality analyses underscored the prognostic impact of cardiopulmonary comorbidities. In the overall cohort, acute myocardial infarction (adjusted HR 1.88), ILD (aHR 1.25), and primary pulmonary hypertension (adjusted HR 1.94) significantly increased mortality risk. These findings align with literature identifying cardiopulmonary involvement as the primary determinant of SSc survival [ 6 , 14 ].

PAH remains a leading cause of SSc-related death [ 6 , 13 – 16 ], with our study confirming its strong lethal association (adjusted HR 1.94). Because PAH often presents with nonspecific symptoms, international guidelines mandate systematic screening. However, specialized diagnostic tools—such as echocardiography [ 17 ], N-terminal pro B-type natriuretic peptide (NT-proBNP) [ 18 ], diffusing capacity for carbon monoxide [ 19 ]—are often unavailable in secondary care. Baseline and symptom-driven screening must be prioritized, with prompt referral to equipped tertiary centers.

Similarly, ILD was associated with increased mortality across all care levels. While immunosuppressive and antifibrotic therapies can stabilize lung function [ 20 ], they cannot reverse established fibrosis. This highlights the critical need for early detection via high-resolution computed tomography (HRCT) [ 21 ]. Limited access to HRCT in secondary hospitals likely drives referral delays, emphasizing the importance of European Respiratory Society (ERS) and the EULAR recommended screening and specialist referral to initiate intervention during the reversible inflammatory stage [ 22 ]. The findings indicate the need for systematic screening for pulmonary hypertension and ILD, as recommended in international guidelines, particularly at the time of diagnosis and early in the disease course [ 17 , 22 ].

Being first diagnosed with SSc in a secondary care hospital was independently associated with higher mortality than first diagnosis in a tertiary care hospital (adjusted HR, 1.28). Several explanations are possible. First, patients initially managed at secondary hospitals may experience delays in receiving comprehensive evaluations for internal organ involvement, advanced imaging, and disease-modifying therapies that are more readily available at tertiary centers. Second, some of these patients may have presented with early, non-specific presentations of SSc, such as the edematous phase characterized by puffy hands or Raynaud’s phenomenon [ 23 ], where the hallmark skin tightness characteristic of the indurative phase is not yet evident, making the initial diagnosis difficult. Third, patients first diagnosed at tertiary centers may represent a subgroup with more easily recognized classical cutaneous features or earlier recognition by primary care physicians [ 5 ], facilitating prompt referral and more timely specialist intervention. These factors likely explain why initial diagnosis at secondary care, even after statistical adjustment, remained a marker of worse prognosis. However, patients initially diagnosed at secondary hospitals may differ systematically from those directly diagnosed at tertiary centers in terms of disease severity, healthcare access, referral practices, geographic factors, and other unmeasured characteristics. Although we adjusted for available baseline characteristics and comorbidities, residual confounding may still remain. Therefore, the observed associations should be interpreted cautiously and not considered causal.

The effect of referral delays among patients first diagnosed at secondary care hospitals further supports the importance of early specialist involvement. The referral time intervals in this study were categorized to reflect both clinical urgency and local administrative mandates. The < 7-day cutoff captures life-threatening emergencies, such as renal crisis or acute cardiac failure, requiring immediate tertiary care. Additionally, the granularity within 30 days aligns with the local Universal Health Coverage (UHC) framework, which requires rapid referral when primary care capabilities are exceeded. While < 30 days is often globally defined as ‘early,’ these tighter thresholds better represent the responsiveness of our regional referral network to acute presentations. Compared with referral within 7 days, all longer referral intervals (8–30, 31–180, and > 180 days) were associated with a significantly increased mortality risk, with adjusted hazard ratios ranging from 1.23 to 1.30. Although the size effects were modest and relatively similar across the delay categories, the direction was consistent: delays beyond the first week were associated with worse survival. This pattern suggests that among patients who reached tertiary care, referral duration was associated with poorer subsequent survival outcomes, such as pulmonary hypertension, severe ILD, or cardiac involvement, which frequently manifest during the early stages of SSc. Rapid skin thickness progression in diffuse cutaneous SSc, for example, is known to be associated with poor outcomes [ 24 ], suggesting that timely intervention afforded by early referral is critical. The necessity of early referral is also shaped by Thailand’s referral system, in which patients under Universal Coverage must initially use primary or secondary care services before specialist referral to tertiary centers is initiated.

Importantly, the robustness of these findings was confirmed through several sensitivity analyses. After excluding patients who died within 30 days of diagnosis at tertiary hospitals—a group that may represent unusually advanced or severe cases—the association between initial diagnosis at secondary care hospitals and increased mortality remained significant, indicating that the observed effect was not solely driven by early deaths or referral of terminal cases. Furthermore, when alternative referral delay thresholds (exceeding 7, 14, 30, and 60 days) were systematically examined, each definition of delayed referral continued to show a significant association with mortality. This consistent pattern across multiple cutoff points strongly reinforces the conclusion that even moderate delays in referral are associated with poorer survival outcomes for SSc patients.

Age has also been consistently implicated as a significant predictor of poor outcomes in patients with SSc. Advanced age has been associated with a higher mortality risk, as shown in registry analyses and meta-analytic data, where older age correlated with worse survival in SSc cohorts [ 6 , 13 , 25 , 26 ]. This partially corroborates our findings that older age was associated with an increased mortality risk among patients first diagnosed at both secondary and tertiary centers. In the time-dependent interaction analysis, mortality risk declined over time, with a 0.5% decrease per month of follow-up (HR 0.995 per month). The observed attenuation of the age effect over time should be interpreted cautiously. This pattern may reflect survivor bias or selective survival, whereby patients with higher early mortality risk die earlier during follow-up, leaving a relatively healthier surviving cohort over time. Furthermore, even when a tertiary center assesses elderly patients, they may still face an adverse prognosis, as the elderly risk profile is compounded by immunosenescence and reduced tolerance for potent immunosuppressive therapies [ 27 , 28 ], often necessitating less aggressive treatment strategies. SSc patients with advanced age should be identified early and referred promptly to specialists for comprehensive evaluation and treatment.

These findings are relevant to Thailand’s healthcare system and other countries with tiered referral structures. The survival disadvantage associated with initial diagnosis at secondary care hospitals and delayed referral points to the need for early recognition and rapid referral pathways. Educational initiatives targeting general practitioners, internists, and physicians in community and general hospitals should focus on early clinical clues and emphasize the need for prompt referral when SSc is suspected. Second, the study’s goal of examining the referral system for SSc highlights the need to develop standardized referral criteria or “red flag” checklists, potentially integrated into electronic health records or referral forms, to improve diagnostic consistency between secondary and tertiary centers. Third, teleconsultation or shared-care models between tertiary rheumatology centers and secondary hospitals may facilitate earlier expert input, particularly in remote or underserved areas, thereby shortening the effective time to specialist assessment, even when physical transfer is delayed.

From a clinical management perspective, our results reinforce the central role of systematic screening and aggressive management of cardiopulmonary complications in patients with SSc. Patients presenting evidence of ILD or pulmonary hypertension, whether identified by ICD-10 codes or clinical evaluation, should be prioritized for early referral and intensive follow-up. In addition, given the association with acute myocardial infarction, attention to cardiovascular risk assessment and management is warranted, including control of traditional risk factors (e.g., hypertension and diabetes mellitus) alongside disease-specific therapy.

To our knowledge, this is the largest nationwide cohort of patients with SSc reported from Thailand, encompassing all patients under the Universal Coverage scheme who were diagnosed with SSc at tertiary or university hospitals over 7 years. The national administrative database allowed us to capture real-world patterns of referral and outcomes across different levels of the healthcare system, providing a comprehensive picture of SSc care beyond single-center experiences. The application of time-to-event methods, including Cox proportional hazards modeling with time-dependent interaction analysis and stratified Cox regression, enabled us to address violations of the proportional hazards assumptions and derive more robust estimates of mortality risk factors. However, several variables violated the proportional hazards assumption, stratified Cox regression was applied. As a result, direct hazard ratio estimates for sex, hypertension, lung cancer, and CKD stage could not be derived from the adjusted model.

However, this study has some important limitations that must be acknowledged. First, the administrative database does not allow identification of the true onset of SSc or total disease duration prior to referral. As a result, the cohort may include both newly diagnosed and previously established cases, which may introduce length-biased sampling. Patients with longer-standing disease who survive to be referred may differ from those with more rapidly progressive disease. This may lead to residual confounding if disease duration or severity influences both referral timing and mortality. The observed associations should be interpreted within the context of the recorded care pathway rather than as causal effects of referral delay. Second, immortal time bias exists because follow-up began at tertiary diagnosis. Patients referred from secondary centers had to survive the referral interval to be included, creating a survivor bias. These likely underestimates mortality in the secondary care group, as those with rapidly fatal disease may have died prior to referral. Thus, the survival benefit in the tertiary group reflects both clinical management and the timing of cohort entry. In addition, patients initially diagnosed at secondary hospitals may also differ systematically from those directly diagnosed at tertiary centers, potentially introducing selection bias and residual confounding that could not be fully addressed using the available administrative data. Third, there is a potential for informative censoring. Although mortality status was cross-referenced with the national death registry, which captures deaths occurring both within and outside hospital settings, some patients may still have been lost to follow-up due to transitions out of the NHSO system. If such loss to follow-up is associated with underlying health status or access to care, it may introduce bias in survival estimates. The direction of this potential bias cannot be determined with the available data, and therefore the findings should be interpreted with caution. Fourth, this study was based on administrative claims data using ICD-10, formal validation using clinical classification criteria was not available. Therefore, some degree of diagnostic misclassification may have occurred, particularly among cases identified only at secondary hospitals. However, diagnoses recorded at tertiary or university hospitals were considered more likely to represent specialist-confirmed SSc. Furthermore, internists working at the tertiary care hospital may recognize the disease and establish the diagnosis during the first visit based on clinical presentation, thereby potentially obviating the need for additional investigations. Consequently, the rate of misdiagnosis or misclassification of SSc in the tertiary hospital may be lower than in secondary centers. Fifth, the NHSO database lacks clinical variables essential for SSc prognostication, including disease subset classification (limited versus diffuse cutaneous SSc), autoantibody profiles (anti-topoisomerase I, anti-centromere, anti-RNA polymerase III), modified Rodnan skin score, pulmonary function parameters, immunosuppressive treatment, socioeconomic status and geographic data. These unmeasured confounders limit our ability to fully adjust for disease severity. Sixth, there was a potential for indication bias, as the reasons for initial presentation at secondary versus tertiary care facilities may differ. We cannot fully capture the specific clinical “reason for visit” that led to the initial encounter at each facility level. Future prospective studies are needed to capture the exact clinical triggers for referral. Finally, socioeconomic status, geographic accessibility to healthcare facilities, travel distance, and institutional resources were not available for analysis and may confound the observed associations between diagnostic site, referral delay, and mortality.

These findings may help implications for healthcare systems worldwide with tiered referral structures and underscore the importance of early specialist evaluation in improving SSc outcomes. This information can be also used to improve the healthcare system and provide better health education for both patients and medical personnel, including medical students and general practitioners, to improve patient outcomes.

Conclusion

In this large nationwide cohort, initial diagnosis at secondary care hospitals and longer recorded referral intervals were associated with poorer survival. Cardiopulmonary comorbidities, particularly primary pulmonary hypertension and acute myocardial infarction, were major risk factors for mortality. Early recognition of SSc in secondary care hospitals and prompt referral to tertiary centers with specialized expertise may contribute to improve patient outcome. Strengthening referral pathways, increasing awareness of early SSc manifestations, and improving access to specialist care may substantially improve survival outcomes in patients with SSc.

Acknowledgements

The authors thank (a) the Scleroderma Research Group and the Faculty of Medicine, Khon Kaen University, for their support; (b) the National Health Security Office (NHSO) for providing access to the database; and, (c) Mr. Bryan Roderick Hamman for assistance with the English-language presentation of the manuscript.

Abbreviations

American College of Rheumatology

Confidence interval

Diffuse cutaneous systemic sclerosis

the European Respiratory Society

the European Alliance of Associations for Rheumatology

Forced vital capacity

Hazard ratio

High-resolution computed tomography

International Statistical Classification of Diseases and Related Health Problems 10th Revision

Interstitial lung disease

Interquartile range

Limited cutaneous systemic sclerosis

the National Health Security Office

N-terminal pro B-type natriuretic peptide

Pulmonary arterial hypertension

Systemic sclerosis

Authors' contributions

Conceptualization: Foocharoen C. Data curation: Foocharoen C, Thinkhamrop W, Onchan T, Pongkulkiat P, Mahakkanukrauh A, Suwannaroj S, and Chaichaya N. Formal analysis: Foocharoen C and Chaichaya N. Funding acquisition: Foocharoen C. Methodology: Foocharoen C, Chaichaya N. Writing – original draft: Foocharoen C. Writing – review & editing: Foocharoen C, Chaichaya N. All authors have read and approved the final manuscript.

Funding

This study was supported by Khon Kaen University, Thailand.

Data availability

The datasets used and/or analyzed in the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The Khon Kaen University Ethics Committee for Human Research determined that this study met the criteria for exemption under the Exemption Determination Regulations ( HE681473 ) on August 6, 2025. This exemption was granted because the study did not involve the collection, use, or analysis of identifiable, private information or biospecimens. As a result, obtaining informed consent from the participants was deemed unnecessary in accordance with the ethical guidelines and regulatory standards for exempt research. The study was conducted in accordance with the principles of the Declaration of Helsinki.

Consent of publication

Not applicable.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

Associated Data

Data Availability Statement

The datasets used and/or analyzed in the current study are available from the corresponding author upon reasonable request.