These data are available from the NHANES database (18). This database comprises a series of nationally representative surveys designed to assess U.S. citizens’ health and nutritional status (19). The database has received approval from the National Center for Health Statistics Ethics Review Board, and the participants have obtained written consent (20). We conducted data analysis on the most recent six NHANES survey cycles (2007–2018), encompassing 15,155 participants, followed by a series of exclusions.

Exclusion of participants with less than 80% of FI features (< 40 items) (n = 244);

Ultimately, it includes 11,234 participants, as depicted in Figure 1.

Following the approach proposed by Hakeem et al., we utilized the FI to assess the degree of frailty. This index comprises 49 variables spanning multiple systems, covering aspects such as cognition, dependency, depressive symptoms, comorbidities, general health status, hospital utilization, physical performance, body measurements, and laboratory test values (21–23). The eligibility survey required a completion rate of at least 80% (approximately 40 items) for the 49 frailty items. We assigned scores ranging from 0 to 1 based on the severity of defects (see Supplementary Table S1)¹. The FI is the sum of defect scores obtained by participants divided by the potential total defect score. With a critical threshold for the FI set at 0.21, values greater than or equal to 0.21 are defined as frail, while values less than 0.21 are non-frail (24).

An automated hematology analyzer will evaluate lymphocyte, neutrophil, platelet, and monocyte counts expressed as ×10³ cells/μL (25). We calculated two systemic inflammatory markers based on peripheral blood cell counts: SII and SIRI. The calculation formula for SII is platelet count × neutrophil/lymphocyte count (26). The calculation formula for SIRI is neutrophil count × monocyte count/lymphocyte count (27).

Correlation logic and previously published literature guided the selection of covariates. We collected statistical data on basic participant information, including age, gender, race, education level, marital status, income poverty ratio, and statistics on alcohol and smoking habits. Specifically, age into two groups: <40 years and ≥ 40 years; gender into two groups: male and female; race into five groups: Mexican American, other Hispanic, non-Hispanic White, non-Hispanic Black, and other races, including multiple races; education level into three groups: less than high school, high school or general education diploma, and more than high school (28); marital status into three groups: married/living with a partner, never married, and separated/divorced/widowed (29); The income poverty ratio is categorized into three groups: low (≤1.3), moderate (1.3–3.5), and high (>3.5). It is calculated by dividing the family’s (or individual’s) income by the poverty threshold specific to the survey year. A lower income poverty ratio indicates lower income for the family (or individual). Based on the response to the question, “Have you smoked at least 100 cigarettes in your entire life?” participants were categorized as smokers or non-smokers. Based on responses to questions about drinking at least 12 alcoholic beverages in the past year and ever drinking any alcoholic beverage, participants were drinkers or non-drinkers (30).

In this study, statistical analyses followed the recommendations of the Centers for Disease Control and Prevention guidelines. We were considering the complexity and multi-stage sampling design of NHANES data collection and sampling weights in the statistical analysis. We compared the baseline characteristics between frail and non-frail individuals, as well as the baseline characteristics of different quartiles of the natural logarithm (ln) transformed SII (lnSII) and SIRI (lnSIRI). For normally distributed quantitative data, we used the t-test, and for qualitative data, we employed the χ² test. Descriptive statistics presented continuous variables using mean ± standard deviation and categorical variables using percentages with 95% confidence intervals. Subsequently, weighted logistic regression models estimate the relationships between lnSII, lnSIRI, and frailty in three different models. Model 1 was a basic unadjusted model; Model 2 adjusted for age, gender, race, income poverty ratio, education level, and marital status; Model 3 included all variables in Model 2, plus smoking and drinking status. Furthermore, restricted cubic splines (RCS) were employed to detect potential non-linear relationships between lnSII, lnSIRI, and frailty. We performed Subgroup and interaction analyses for age, gender, education, marital status, income and poverty ratio, smoking, and drinking. All analyses use R software (V.4.3.2), Stata software (version 17), and SPSS software (version 27). Statistically significant: A significance level of p < 0.05 was considered.

11,234 participants were included, with males comprising 45.95% and females 54.05%. Clinical characteristics of participants, stratified by frailty status, are presented in Table 1. The proportions of lnSII four quartiles among frail patients were 21.24, 21.80, 23.7, and 33.26%, respectively, while lnSIRI four quartiles were 20.08, 21.25, 25.01, and 33.67%, respectively. Frailty showed statistical significance (p < 0.05) concerning age, gender, race, education level, marital status, income poverty ratio, alcohol consumption, smoking status, lnSIRI, and lnSII. Compared to non-frail individuals, frail patients are often female, have lower educational levels, lower income poverty ratio, are less likely to be married or living with a partner, and are more likely to smoke. Additionally, they tend to have higher levels of lnSII and lnSIRI. The mean age of frail patients is 64.29 ± 0.22 years.

Among the 11,234 participants, 63.54% were non-frail, and 36.46% as frail. Stratifying by different levels of lnSIRI, significant differences in frailty, age, gender, race, education, marital status, income and poverty ratio, alcohol consumption, and smoking status were observed (p < 0.05). The proportions of the four categories of frailty were 34.96, 29.48, 34.43, and 46.69%, as illustrated in Table 2.

Stratifying by different levels of lnSII, significant differences in frailty, race, education, marital status, income and poverty ratio, and smoking status were observed (p < 0.05). The proportions of the four frailty categories were 35.91, 31.12, 33.05, and 45.35%, as illustrated in Table 3.

Three weighted logistic regression models were employed to investigate the association between lnSII, lnSIRI, and frailty. The odds ratios (OR) and 95% confidence intervals (CI) for the ratio of frailty’s correlation with lnSII and lnSIRI are presented in Table 4. Model 1, unadjusted; Model 2, adjusted for age, gender, race, education, marital status, income-to-poverty ratio; Model 3, additional adjustment for smoking and drinking status based on Model 2. There is a consistently significant positive correlation between lnSII, lnSIRI, and frailty in all three models. The impact of lnSII on frailty was as follows: Model 1 (OR = 1.34, 95% CI: 1.21–1.48), Model 2 (OR = 1.41, 95% CI: 1.26–1.57), Model 3 (OR = 1.38, 95% CI: 1.24–1.54). When assessing different levels of lnSII, compared to the second quartile of lnSII, the first and third quartiles did not show a significant difference in the probability of frailty. At the same time, the likelihood significantly increased in the fourth quartile: Model 1 (OR = 1.84, 95% CI: 1.58–2.13), Model 2 (OR = 1.86, 95% CI: 1.59–2.17), Model 3 (OR = 1.82, 95% CI: 1.55–2.12). Furthermore, the trend p-values for all three models were below 0.001.

The impact of lnSIRI on frailty was as follows: Model 1 (OR = 1.45, 95% CI: 1.32–1.58), Model 2 (OR = 1.73, 95% CI: 1.56–1.91), and Model 3 (OR = 1.69, 95% CI: 1.53–1.88). When assessing different levels of lnSIRI, compared to the second quartile of lnSIRI, the first quartile did not show a significant difference in the probability of frailty. In contrast, the probabilities significantly increased in the third and fourth quartiles. For the third quartile: Model 1 (OR = 1.26, 95% CI: 1.08–1.47), Model 2 (OR = 1.32, 95% CI: 1.12–1.55), Model 3 (OR = 1.30, 95% CI: 1.10–1.53); For the fourth quartile: Model 1 (OR = 2.09, 95% CI: 1.80–2.44), Model 2 (OR = 2.34, 95% CI: 1.99–2.76), Model 3 (OR = 2.29, 95% CI: 1.95–2.70). Furthermore, the trend p-values for all three models were below 0.001.

We also used RCS to visualize the association between lnSII, lnSIRI, and frailty. After adjusting for all covariates in the primary analysis Model 3 mentioned above, we observed a non-linear correlation between lnSII, lnSIRI, and frailty (Figure 2).

The interaction tests in the subgroup analysis revealed that the impact of lnSIRI on frailty varied with age (p for interaction = 0.007) and income-to-poverty ratio (p for interaction <0.001). However, gender, race, education, marital status, smoking, and alcohol consumption did not influence this positive correlation (p for interaction >0.05), as illustrated in Figure 3.

The relationship between lnSII and frailty showed no statistically significant differences across different strata, indicating that age, gender, race, education, marital status, income-to-poverty ratio, smoking, and alcohol consumption did not significantly affect this positive correlation (p for interaction >0.05), as depicted in Figure 4.