Taiwan’s National Health Insurance (NHI) is a nationwide medical financial support system commenced in 1995.12) Over 99% of Taiwanese residents, approximately 23 million people, benefit from this system. The National Health Insurance Research Database (NHIRD) is derived from NHI and collects medical activities, including demographic data, disease diagnosis, prescriptions, surgeries, and expenses. This retrospective cohort study utilized outpatient and inpatient data obtained from the NHIRD. Participants were selected based on the codes from the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) [13,14]. The database review was conducted by the Management Office of Health Data at China Medical University Hospital’s Clinical Trial Center. Rickets and associated variables were categorized based on clinical relevance and coded for statistical analysis using corresponding ICD codes. This classification was reviewed periodically by senior members of the research team for quality assurance.

The study was conducted anonymously to protect the privacy, rights, and interests of the participants. The NHIRD provided deidentified data, ensuring the concealment of identifiable personal information of the patients. As all identifying personal information was removed from the secondary files before analysis, patient consent was not required for accessing the NHIRD. However, some analyzed data were absent due to government regulations prohibiting transportation if the data variable unit is less than 3 units. This precaution aims to prevent the identification of extremely rare cases through alternative system analyses. All protocols were approved by the Institutional Review Board of the Research Ethics Committee of China Medical University Hospital (CMUH 110-REC3-133(CR-2)).

From January 1, 2008, to December 31, 2018, a total of 31,488,321 outpatients and inpatients were recorded in the NHIRD. Detailed information regarding the ICD-9-CM and ICD-10-CM codes used in the study is provided in Supplementary Table 1. Data from the “detailed documents of hospitalization medical expenses” and “registry for contracted medical facilities” were extracted from the NHIRD. The index date was defined as the date of the initial diagnosis of rickets for each patient. We identified a total of 10,276 patients with a diagnosis of nutritional rickets or hereditary rickets (nutritional rickets ICD-9: 268.0, 268.1, 268.9; ICD-10: E55.0, E55.9, E64.3; hereditary rickets ICD-9: 275.3; ICD-10: E83.30, E83.31, E83.32, E83.39). After excluding patients whose information was incompatible with rickets based on the timeframe before 2008 and after 2018 (n=1,795), those aged over 18 years (n=6,639), and those with unknown sex (n=291), we finally analyzed 1,551 participants with rickets (Supplementary Fig. 1). In the NHIRD, sex may be masked in specific cases (unknown sex) to comply with anonymization policies, particularly when subgroup sizes are small and pose potential re-identification risk.

We examined the sociodemographic factors, including sex, age, monthly household income, season of disease onset, urbanization, and level of residence based on disease complexity. Patients were classified into subgroups according to these covariates: preschoolers (age 0–5), school-age children (age 6–10), and adolescents (age 11–18) by age; low income (New Taiwan dollar [NTD]<20,000), medium income (NTD 20,000–39,999), and high income (NTD≥40,000 by monthly household income; and urbanization levels ranging from 0 (low urbanization) to 3 (high urbanization).

The baseline comorbidity history included prematurity, low birth weight, underweight, anemia, enthesopathy, osteoarthrosis, chronic kidney disease (CKD), hyperparathyroidism secondary to renal tubulopathies, and infectious lung diseases (Supplementary Table 1). These comorbidities were included as categorical variables in the models to investigate their impact on the mortality in patients with rickets.

We examined the overall incidence and mortality rate of rickets annually from 2008 to 2018. The distributions of sociodemographic factors, including age, season of disease diagnosis, and income levels, were analyzed. The annual incidence of rickets patients in the overall group and sex subgroups was investigated. Additionally, potential factors influencing the ROM were thoroughly evaluated among both survival and mortality groups. The length of hospital stay (LOS) among rickets patients was also examined to elucidate its correlation with the mortality rate.

For descriptive statistics, we used the chi-square test or Fisher exact test to analyze differences in categorical variables, and one-way analysis of variance for continuous variables. These tests were used to compare demographic characteristics and comorbidities between all participants and the mortality group, as shown in Table 1.

To identify independent factors associated with mortality, the Cox proportional hazards regression model was performed, incorporating relevant demographic and healthcare-related covariates described above (Table 2). Data are expressed as adjusted hazard ratios (HRs) with 95% confidence intervals (CI). All statistical analyses were performed using IBM SPSS Statistics ver. 22.0 (IBM Co., USA), with a 2-tailed P value<0.05 indicating statistical significance.

Among the 1,551 index patients, male comprised 52.87% and females 47.13% (Table 1). The mean age was 6.44±6.09 years. Preschool-aged children represented the largest group (50.61%), followed by adolescents (34.04%) and school-aged children (15.34%). Most patients came from middle-income households (56.54%). While the occurrence of rickets did not vary significantly across seasons, and a high proportion of patients resided in areas characterized by higher levels of urbanization (level II: 32.17%, level III: 49.32%).

A stratified analysis of all participants into nutritional rickets and hereditary rickets categories revealed twice as many cases in nutritional rickets compared to hereditary rickets (1,006 cases vs. 545 cases). Nutritional rickets showed an equal sex distribution, whereas hereditary rickets had a male-female ratio of 3:2. The mean age was younger in nutritional rickets (5.61±5.72 years vs. 7.97±6.46 years). Preschoolers predominated in nutritional rickets (55.57%), while adolescents were more prevalent in hereditary rickets (43.49%), suggesting a potential delay in diagnosing hereditary rickets. Family income, seasonal onset, and urbanization level were comparable across both categories.

We further analyzed the demographic characteristics and associated comorbidities among the mortality group. Male rickets patients (56.76%) and preschoolers (41.89%) exhibited a high tendency toward mortality (Table 1). A low family income level (<20,000 USD: 45.95%) was also noted among the mortalities. Additionally, the mortality group had a higher proportion of patients with anemia (mortalities: 37.84% vs. total population: 10.35%), CKD (mortalities: 9.46% vs. total population: 3.29%), and hyperparathyroidism secondary to renal tubulopathies (mortalities: 9.46% vs. total population: 2.13%). Notably, the mortality group had a significantly longer LOS >7 days (77.03%, P<0.001), with a mean LOS of 28.16±51.4 days. Similar findings were observed when conducting a stratified analysis based on nutritional or hereditary rickets.

The Cox proportional hazards regression model was employed to identify influential factors for rickets mortality (Table 2). After adjusting for variables, sex, age, season of diagnosis, and urbanization level did not significantly affect the ROM. However, patients with medium and high incomes had a significantly decreased ROM compared to those with low incomes (0.13-fold, P<0.001; 0.27-fold, P<0.001). Using the receiver operating characteristic (ROC) curve, we identified the optimal cutoff value of LOS for predicting mortality as 3 days for overall rickets (Supplementary Table 2). Furthermore, patients with a longer LOS (>3 days) had a 4.82-fold increased ROM compared to those without (P<0.001). Similar findings were observed in a stratified analysis of nutritional or hereditary rickets (Table 2).

Supplementary Table 3 displays the trend of rickets occurrence among patients during the follow-up period. Overall, the incidence showed a mild decrease from 2008 (2.66 per 105 population) to 2011 (1.41 per 105 population), followed by an inflection point in 2012 (1.83 per 105 population), and an increasing trend thereafter (Fig. 1A). Additionally, the annual incidence in both sexes was similar to that of the overall population. The annual percentage change (APC) was significantly noted in total population (13.56, P=0.002) and both sexes (males=13.09, P=0.004; females=14.55, P=0.002). Regarding subgroups analysis, we found that the trend of the incidence rate in nutritional rickets (APC=17.9, P=0.001) aligned with that of overall rickets, while hereditary rickets exhibited a steady low incidence (APC=5.73, P=0.028) (Fig. 1B and C).

The highest mortality rate among index patients occurred in 2008 (males, 17.65%; females, 8.11%) (Fig. 2A). Overall, there was a significant decreasing trend in mortality rates for rickets (-13.48, P=0.002), though there was a slight increase in males and a slight decrease in females (males=-3.50, P=0.942; females=-48.18, P=0.353) (Fig. 2A). A slight decrease in mortality rates was observed among the total population with nutritional rickets (APC=-46.20, P=0.307) (Fig. 2B). Similarly, there was also a slight decline among the total population with hereditary rickets (APC=-4.86, P=0.293), although its trend pattern appears to be influenced by the results for males (Fig. 2C). The detailed case numbers of annual mortality in index cases from 2008 to 2018 was shown in Supplementary Table 4.

LOS was assessed to evaluate rickets severity and its correlation with mortality. Patients with hereditary rickets had significantly longer mean LOS compared to those with nutritional rickets (11.77"±" 19.6 vs. 3.51"±" 15.89, P<0.001) (Table 1). The mortality group showed a higher mean LOS compared to the overall group (28.16±51.4 vs. 6.41±17.72, P<0.001). Notably, mortality cases in nutritional rickets had a significant longer mean LOS than those in hereditary rickets (32.32±79.9 vs. 26.4±33.79, P<0.001). The area under the ROC curves was 0.8085 for the overall group (95% CI, 0.760–0.857; P<0.001), 0.7414 for the nutritional groups (95% CI, 0.661–0.888; P<0.001), and 0.720 for the hereditary group (95% CI, 0.653–0.787; P<0.001) (Fig. 3).

To our knowledge, there have been limited epidemiological studies on rickets, primarily focusing on the Caucasian population [15,16]. Our large-scale nationwide retrospective study aims to fill this gap by investigating the Asian population affected by nutritional or hereditary rickets and providing comprehensive epidemiological data, including factors influencing ROM and annual incidence [6,8,9,17-21]. In Taiwan, a developed country, the incidence of rickets increased from 2012 to 2018, potentially due to improved diagnostic capabilities rather than solely nutritional deficits. Overall, rickets slightly predominates in males and manifests during preschool or adolescence. ROM is linked to low income, anemia, CKD, hyperparathyroidism secondary to renal tubulopathies, and extended LOS. Early recognition and management of these factors are essential to avoid severe outcomes.

The comparison of significant findings between the current study and previous studies is summarized in Table 3. Wheeler et al. [8] conducted a prospective cohort study to elucidate the temporal relationship between nutritional rickets and risk factors such as darker skin pigmentation, Indian and African ethnicity, age under 3 years, and exclusive breastfeeding; however, the limited number of patients (n=58) from a specific region might restrict generalizability. Apart from 2 studies on X-linked hypophosphatemia [17,18], most retrospective studies on nutritional rickets have identified factors such as black race, exclusive breastfeeding, low birth weight, males, lower socioeconomic status, lack of sun exposure, and poor nutritional condition as increasing the risks of rickets [6,20,21]. Notably, our study simultaneously investigated both hereditary and nutritional rickets, providing a broader perspective on the care of rickets. Unlike other studies with small sample sizes [9,19], our nationwide retrospective study (1,551 cases) revealed a notable prevalence of rickets among pediatric patients, with a higher proportion diagnosed with nutritional compared to hereditary rickets. Our findings also suggest an increased overall incidence of rickets alongside declining mortality rates in recent years. However, these findings were limited by the lack of anthropometric data and detailed biochemical analysis. Future prospective research with larger sample sizes and comprehensive data is necessary to better understand the clinical characteristics and risk factors of rickets in diverse populations.

Previous studies have reported a male predominance in rickets incidence [22,23]; however, our study offers a more nuanced perspective. While males outnumber females among rickets patients overall, subgroup analysis reveals that only hereditary rickets exhibit a male predominance, whereas nutritional rickets show a nearly equal sex distribution. Although the discrepancy between our findings and previous reports may be influenced by racial and genetic factors [22,23], our results appear to better align with biological plausibility. Nutritional rickets primarily arises from individual nutritional status, indicating no inherent sex difference under similar environmental conditions [24]. Conversely, X-linked hypophosphatemia, a leading cause of hereditary rickets, may account for the higher likelihood of males being affected in this subgroup [25].

Hereditary rickets is often diagnosed later, typically during adolescence. This delayed diagnosis reflected the real-world challenges in prompt diagnosis in Taiwan. Several factors may contribute to this delay: the clinical complexity and subclinical nature of rickets can lead to under-recognition or misdiagnosis until symptoms become more pronounced in later childhood or adolescence [26]. The limited use of genetic analysis in Taiwan might result in patients consulting multiple healthcare providers before receiving a confirmed diagnosis [27]. Additionally, restricted access to pediatric endocrinologists specializing in rickets contributed to diagnostic delays. Finally, familial variability in disease presentation [28] and a cultural reluctance among Taiwanese parents to seek medical care due to feelings of guilt or shyness may further delay diagnosis.

In developing countries, the higher prevalence of rickets is often attributed to malnutrition stemming from poorer socioeconomic conditions [29]. However, our findings show that the most patients come from middle-income families with annual incomes exceeding 20,000 USD. This may be due to Taiwan’s status as a developed country with relatively small income disparities [30]. Additionally, the majority of rickets-related deaths occurred in low-income families, aligning with previous studies that report increased rickets prevalence in developed countries, possibly linked to immigration or refugees issues that restrict access to nutritious food and healthcare resources. Consequently, higher incidences and increased severity of rickets cases ensue [20,31].

Our study indicates that nutritional rickets generally presents with lower severity, evidenced by a shorter-than-expected average LOS compared to hereditary rickets. This suggests challenges in detecting and diagnosing hereditary rickets in Taiwan [32]. While nutritional rickets can often be resolved by improving nutritional status, hereditary rickets requires lifelong management due to genetic deficits [5,33,34]. Furthermore, our findings suggest that rickets patients with high disease severity face increased mortality risks, often associated with comorbidities such as anemia, CKD, and hyperparathyroidism secondary to renal tubulopathies. Possible explanations for these associations include the relationship between vitamin D deficiency and pulmonary infections like pneumonia, as well as the anemia’s interference with treatment efficacy, leading to higher mortality rates among severely ill patients [35]. Additionally, complications related to CKD, such as bicarbonate wasting, alkaline urine, and renal tubular acidosis, can impede rickets treatment, potentially resulting in refractory cases [36]. Severe nutritional rickets may also contribute to secondary hyperparathyroidism [37,38], increasing the risks of osteoporotic fractures and mortality due to disruptions in calcium and phosphorus metabolism [39]. Nonetheless, deaths in rickets are rarely a direct result of bone instability; they are usually caused by associated comorbidities. To effectively reduce mortality, precise interventions should prioritize improving these main contributing comorbidities.

From 2012 to 2018, the overall incidence of rickets in Taiwan showed a steady annual increase, largely driven by nutritional rickets rather than the stable occurrence of hereditary forms. Several factors may explain the rising incidence of nutritional rickets. The measurement of 25-hydroxyvitamin D levels has become more widespread in recent years [40]. In Taiwan, milk sold in containers is not fortified with vitamin D, and health supplements containing vitamin D are not commonly used by the general population. Given that over half of mothers and newborns were vitamin D deficient at birth [41], another possible reason could be the increase in breastfeeding prevalence in Taiwan, which rose from 32.2% in 2008 to 49.9% in 2011 and continues to rise [42]. In 2012, the Taiwan government intensified efforts to promote breastfeeding through initiatives like advocating exclusive breastfeeding in maternity hospitals and establishing breastfeeding-friendly facilities in public areas. While exclusive breastfeeding offers significant benefits, it has also contributed to increased vitamin D deficiency prevalence [43]. Consequently, the Taiwanese government recently endorsed additional vitamin D supplementation, aligning with international recommendations [44].

Beyond its statistical significance, prolonged hospitalization may also hold clinical implications as a red flag indicator for adverse outcomes. Our ROC analysis showed that a LOS >3 days was the optimal threshold for predicting mortality, with a sensitivity of 82.4% and specificity of 68.9% (Supplementary Table 2), and an AUC of 0.81 (Fig. 3). These findings are consistent with previous studies that demonstrated a positive association between extended LOS and mortality in pediatric patients [45]. In clinical settings, children with rickets who remain hospitalized beyond 3 days may warrant prompt multidisciplinary reassessment, including evaluation of comorbidities, nutritional status, or potential delays in treatment response. Nevertheless, LOS can be influenced by nonmedical factors such as discharge planning or resource availability, which may limit its specificity. Additionally, the possibility of reverse causality cannot be excluded—children with higher baseline severity or greater risk of death may have required prolonged hospitalization, rather than extended LOS being the cause of poor outcomes. Thus, further prospective research is still needed to validate its utility as a clinical indicator of elevated mortality risk in rickets.

This study has several limitations. Firstly, the NHIRD did not provide comprehensive information such as body height, weight, body mass index, precise dosage of rickets medications, primary cause for admission, some socioeconomic details (e.g., educational level), and environmental factors (e.g., nutritional status). Secondly, biochemistry and endocrine parameters (e.g., serum electrolytes, parathyroid hormone, 25-hydroxyvitamin D, and 1,25-dihydroxyvitamin D) were unavailable in the NHIRD. Furthermore, given the absence of genetic testing data in the NHIRD, hereditary rickets was defined exclusively using ICD diagnostic codes for disorders of phosphate metabolism (ICD-9, 275.3; ICD-10, E83.30–E83.39). This methodology may result in misclassification bias, as genetic confirmation was not available for all patients and the diagnostic codes may have inadvertently included cases of secondary metabolic bone disorders, acquired malabsorption disease, nutritional deficiencies, and renal tubular disorders. Thirdly, imaging studies like bone surveys were not included, potentially affecting outcome analysis. Lastly, the outcomes of rickets treatment were not evaluated due to limitations in assessing intervention implications from NHIRD data. Despite these limitations, the NHIRD provides comprehensive population coverage, reducing the potential for recall and selection bias. However, further prospective studies with more detailed information are necessary to fully elucidate the impact of rickets and its treatments on normal growth and ROM.

In conclusion, our study highlights the demographic characteristics, ROM, and temporal trends in incidence and mortality among patients with rickets. We found that rickets primarily affects preschool-aged children. Hereditary rickets is less common than nutritional rickets and often diagnosed later, typically in adolescence. Significant mortality predictors include low family income and longer LOS. Despite rising incidence rates, mortality rates have decreased, underscoring the importance of targeted interventions and monitoring strategies to mitigate the burden of rickets and improve outcomes.