The National Health and Nutrition Examination Survey (NHANES) is a comprehensive national probability sample of the entire non-institutionalized civilian population in the USA, covering individuals aged 2 months and above. The survey collected a wide array of data, such as demographic information, socioeconomic status, dietary behaviors, and various health-related details, and involved thorough examinations and laboratory tests administered by highly skilled medical professionals. NHANES data are publicly accessible through the NHANES website.¹ Our study utilized data from six NHANES cycles spanning 2007–2018. The surveys have been approved by the NCHS Institutional Review Board, and participants have provided informed written consent. The study focused on adults aged 20 years and above who completed the survey. Subjects without relevant data on kidney stones, mortality, or DSI were systematically excluded, resulting in a final sample of 2,765 subjects (Figure 1).

In NHANES, a 24-h recall was used to inquire about all foods and beverages consumed from midnight to midnight on the day preceding the interview. Nutrient data associated with the consumption were sourced from the United States Department of Agriculture’s Food and Nutrient Database for Dietary Studies. The average sodium intake (mg/day) was calculated using dietary recalls from each survey period. The recommended sodium intake was defined as being below 2,300 mg/day. Referring to the Dietary Guidelines for Americans 2020–2025 (9), DSI was categorized into four groups (<2.2 g/d, 2.2–3.1 g/d, 3.1–4.2 g/d, >4.2 g/d).

Data on the prevalence of kidney stone disease were obtained from the interview records. Participants were asked, “Have you ever experienced a kidney stone?” during the standardized home interview. Adult participants who answered affirmatively were categorized as having a history of kidney stones.

Data on mortality status and follow-up duration (in months) were collected from death certificates in the National Death Index (NDI), covering the period from survey participation to December 31, 2018. Our study’s primary findings focused on mortality from all causes.

Demographic and socioeconomic data, such as age, gender, race/ethnicity, education level, marital status, poverty income ratio (PIR), and smoking status, were collected through structured household interviews using a standardized questionnaire. In addition, height, weight, and alcohol consumption were measured during the participants’ visits to mobile health screening centers for medical assessments. Body mass index (BMI) was computed according to a standardized procedure by dividing the weight in kilograms by the square of height in meters (10). Race/ethnicity categories comprised Mexican American, Non-Hispanic White, Non-Hispanic Black, and other races. Education levels were classified as less than high school, high school diploma or equivalent, and college education or higher. Marital status was categorized as ‘married’ and ‘not married’, encompassing individuals who had never been married, those living with a partner, and others. Family income was determined based on the poverty income ratio (PIR), which considered specific thresholds according to family size (≤1.3, 1.3–3.5, >3.5). Smoking status was classified into three groups: “never smoker” for those who had never smoked cigarettes, “former smoker” for individuals who had smoked ≥100 cigarettes in the past but had quit, and “current smoker” for those who had previously smoked ≥100 cigarettes and were still smoking at the time of the interview (11). Drinking status was determined by whether the individual consumed 12 or more alcoholic beverages per year (12). A self-reported medical history of diabetes, hypertension, cardiovascular disease, and stroke was also collected.

Histogram distribution, Q–Q plot, or the Kolmogorov–Smirnov test was used to determine whether variables conformed to a normal distribution. All normally distributed continuous variables were presented as mean ± SD, while skewed continuous variables were described as median (interquartile range [IQR]). Categorical variables were presented as frequencies (%). The chi-square (categorical variables), One-Way ANOVA test (normal distribution), or Kruskal-Wallis H test (skewed distribution) was employed to analyze differences among different DSI groups.

Hazard ratios (HRs) and 95% CIs were calculated to determine the association between all-cause mortality and DSI by using Cox proportional hazards models. Participants who were lost to follow-up were censored at that particular time point. The outcome of all-cause mortality was assessed with Kaplan–Meier survival curves according to the DSI (as a categorical variable) and evaluated with the log-rank test. Confounders were selected based on their associations with the outcomes of interest, or a change in the effect estimate of more than 10%. Specifically, four models were constructed: (1) unadjusted; (2) adjusted for age and gender; (3) adjusted for age, gender, and race; and (4) adjusted for age, gender, race, BMI, hypertension, diabetes mellitus, coronary heart disease, stroke, smoking, and alcohol consumption. Subgroup analyses were performed to assess the link between DSI and all-cause mortality, with interaction p-values tested within each subgroup. Weighted generalized additive model regression and smoothed curve fitting were employed to delve deeper into the association between DSI and all-cause mortality. The calculated effect sizes and p values from all these models were reported and compared. All analyses were performed using R Statistical Software (Version 4.2.2, The R Foundation)² and Free Statistics analysis platform (Version 1.9, Beijing, China).³ FreeStatistics is a software package that provides intuitive interfaces for most common analyses and data visualization. The software uses R as the underlying statistical engine, and the graphical user interface (GUI) is written in Python. This package is designed for reproducible analysis and interactive computing. In this study, a two-sided p value < 0.05 was considered statistically significant.

Among the 2,765 adults diagnosed with kidney stone disease, the median age was 56.1 years (25th percentile: 48 years, 75th percentile: 64 years), with 55.57% being male. Within this cohort, 25.0% had a low DSI (<2.2 g/day). Table 1 presents the weighted sociodemographic characteristics and various disease conditions across the four groups. Individuals with higher DSI tended to be younger, with a higher proportion of Non-Hispanic White people, higher educational attainment, stable marital status, higher household income, lower prevalence of coronary heart disease (CHD), and higher likelihood of smoking and alcohol consumption compared to participants with lower sodium intake (<2.2 g/d) (p < 0.01). However, no significant difference in body mass index (BMI), hypertension, diabetes mellitus (DM), and stroke was found among the four groups (all p value > 0.05).

In the cohort of 2,765 patients, all-cause mortality showed an incidence rate of 13.5%. Univariate analysis revealed a significant inverse association between DSI and mortality, as evidenced by the Kaplan–Meier curves (log-rank test p = 0.027, Figure 2). The multivariable Cox proportional hazards analysis (Table 2) indicated that higher sodium intake was associated with a reduced mortality risk (per 1,000 units: HR = 0.89, 95% CI = 0.80–0.99, p = 0.034) in model 4, which was adjusted for confounders. The results suggested an 11% decrease in the risk of all-cause mortality for each additional g/day increase in DSI. Adjusted hazard ratios (95% CIs) for mortality in sodium intake groups 2, 3, and 4 (2.2–3.1 g/d, 3.1–4.2 g/d, >4.2 g/d) were 0.78 (95% CI = 0.54, 1.13), 0.66 (95% CI = 0.46, 0.96), and 0.70 (95% CI = 0.44, 1.13) compared to group 1 (<2.2 g/d) (Table 2, model 4). The model shows that participants in groups 2, 3, and 4 had a 22, 34, and 30% reduction in mortality incidence, respectively, compared to group 1 (Table 2, model 4). As depicted in Figure 3, the interaction between DSI and all-cause mortality was analyzed in participants subgroup by age (<60 or ≥60 years), gender (female or male), BMI (<24, 24–28, or ≥28 kg/m²), smoking status (never smoker, former smoker, or current smoker), drinking status (no or yes), race (Non-Hispanic White or others), education level (<high school, high school, >high school), and PIR (<1.3, 1.3–3.5, >3.5). No significant interactions were observed between DSI and other stratified variables (p for interaction >0.05). Figure 4 depicts the dose–response correlation between DSI and all-cause mortality. Following comprehensive adjustments, a linear correlation was observed (p for nonlinearity > 0.05). Each increase of 1,000 units (g/d) in DSI correlated with an 11% decrease in the risk of all-cause mortality (refer to Table 2).