The non-linear association between depression scores and all-cause mortality: a cohort study based on NHANES 2005–2018 data

Abstract

Depression is an important public health problem and its association with mortality has been studied extensively. However, the relationship between different levels of depression and death in adults is not well understood. This study aimed to explore the association between depression scores and all-cause mortality in US adults. We conducted a retrospective cohort study using data from 2005 to 2018 National Health and Nutrition Examination Survey (NHANES). Depression was assessed using the Patient Health Questionnaire-9 (PHQ-9), and all-cause mortality was the primary outcome. Cox proportional hazards models were used to examine the association between depression scores and mortality. A two-piece wise linear regression model was used to examine the threshold effect. A total of 36,393 participants with a mean age of 47.9 years (SD = 18.8) were included. The median follow-up time was 89 months, during which time 3,644 (10.01%) deaths occurred. When the depression score was below 7, each unit increase in the score was associated with a 6% increased in the risk of all-cause mortality (HR = 1.06, 95% CI: 1.05–1.08, P < 0.0001). The results of this study show a non-linear association between depression scores and all-cause mortality among adults in the United States. Increased depression scores were associated with increased mortality. However, these findings need to be further validated by further research.

Depression is recognized as a leading cause of disability globally and is significantly linked to increased mortality risk. A meta-analysis indicated that depression is associated with increased risk of CVD mortality^1,2. Furthermore, research has shown that the severity of depressive symptoms correlates with the risk of coronary heart disease and stroke^3,4, suggesting a complex relationship where each incremental increase in depressive symptoms can significantly elevate mortality risk^5,5. Some studies suggest that the relationship may not be straightforward, as the risk of mortality appears to be influenced by the presence of comorbid conditions^6,7. For example, individuals with depression and multiple comorbidities face the highest mortality risk, yet they also tend to benefit the most from targeted depression interventions⁸.

The relationship between depression and mortality has been a subject of substantial academic inquiry, yielding various models that illustrate its complexities. Research indicates that higher levels of depression are correlated with increased mortality rates. For instance, a comprehensive study affirms that depression is a critical risk factor for multiple disease-related causes of death, extending beyond cardiovascular disease and suicide⁹. However, this linear perspective has been challenged by findings suggesting more nuanced relationships. Other studies propose that the relationship between depression and mortality may not be strictly linear but rather follow nonlinear patterns, such as U-shaped or J-shaped curves. For example, research demonstrates a U-shaped trajectory of depression among older adults, indicating that both low and high levels of depression correlate with higher mortality rates, while moderate levels may not significantly elevate mortality risk¹⁰. Further supporting this nonlinear view, research indicates that the detrimental effects of depression on mortality vary significantly depending on severity. Severe depression has been linked with considerably elevated mortality risks compared to individuals with moderate depressive symptoms, who may not face a significantly higher risk¹¹. These findings underline the importance of considering the severity of depressive symptoms when assessing mortality risks. This complexity underscores the need for a better understanding, which is vital for developing targeted interventions to reduce mortality risks in depressed individuals^12,13.

The National Health and Nutrition Examination Survey (NHANES) provides a unique opportunity to explore this relationship in a large, nationally representative sample. The present study aims to investigate the association between depression scores, as measured by the PHQ-9, and all-cause mortality.

Methods

Data source and study population

We used data from the NHANES 2005–2018 cycles, which include detailed health and nutrition information from a nationally representative sample of the U.S. population¹⁴. Participants aged 18 years and older with complete data on depression scores and mortality status were included. The study flowchart was presented in Fig. 1. The final sample consisted of 36,393 participants. The NHANES study was approved by the National Center for Health Statistics (NCHS) Research Ethics Review Board, and all participants provided written informed consent. The data used in this study are publicly available from the NHANES database (https://www.cdc.gov/nchs/nhanes/index.htm) and the National Death Index (https://www.cdc.gov/nchs/data-linkage/mortality-public.htm).

Fig. 1.

Flow chart of study population. NHANES, National Health and Nutrition Examination Survey.

Depression assessment

Depression was assessed using the Patient Health Questionnaire 9-Item (PHQ-9)¹⁵, a validated 9-item questionnaire that scores depression severity from 0 to 27. The presence of clinically significant depressive symptoms was determined using the established threshold of a PHQ-9 score ≥ 10. For most analyses, participants were further categorized into no depression (PHQ-9 score: 0–4), mild (PHQ-9 score: 5–9), moderate (PHQ-9 score: 10–14) and severe depression (PHQ-9 score ≥ 15)¹⁶.

Outcome

The primary outcome was all-cause mortality, with follow-up data obtained through linkage to the National Death Index (NDI)¹⁶ through December 31, 2019. Cardiovascular disease (CVD) mortality is determined according to the guidelines of the Tenth Revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10), including (1) diseases of heart (I00-I09, I11, I13, I20-I51); (2) cerebrovascular diseases (I60-I69).

Covariates

At baseline, all participants completed in-person interviews and physical examinations. Demographic data (e.g., age, sex, race/ethnicity, education), lifestyle factors (e.g., smoking, alcohol use), and physical measurements were collected. BMI was categorized per CDC guidelines as underweight (< 18.5 kg/m²), normal (18.5–24.9 kg/m²), or overweight (≥ 25 kg/m²)^17,18. Socioeconomic status was assessed using the Poverty Income Ratio (PIR), calculated as the ratio of family income to the poverty threshold, with values grouped as poor (< 1.0), near poor (1.0-1.9), middle income (2.0-3.9), and high income (≥ 4.0). Smoking status was defined as never, former, or current. Alcohol intake was based on self-reported frequency and quantity over the past year and categorized as 0, 1, 2, or ≥ 3 drinks/day¹⁹. Physical activity was assessed by self-reported duration and intensity of recreational activities, with weekly MET-minutes calculated per NHANES protocols²⁰. Participants were classified as physically active if meeting ACSM guidelines (≥ 150 min/week of moderate or ≥ 75 min/week of vigorous activity)²¹. Hypertension was defined when the participants self-reported hypertension or used medication for hypertension. For those who did not have self-reported, according to the American Heart Association/American College of Cardiology (AHA/ACC) 2017 guideline for monitoring and diagnosis of hypertension. Participants with systolic BP ≥ 130 mmHg or diastolic BP ≥ 80 mmHg were also considered hypertensive²². Diabetes was defined by self-reported diagnosis, fasting glucose ≥ 126 mg/dL, or HbA1c ≥ 6.5% ²³.

Statistical analysis

We first describe the baseline characteristics of the study population. Continuous variables were expressed as mean ± standard deviation, categorical variables were expressed as weighted percentages, and differences were compared in groups based on depression.

In the main effects analysis, Cox proportional hazard regression model was used to evaluate the association of baseline PHQ-9 depression score on the risk of all-cause death. The hazard ratio (HR) and 95% confidence interval (CI) were calculated with survival time as the horizontal axis and death event as the end point. Survival time was calculated as the number of months from the date of the NHANES investigation to the date of death or the end date of follow-up (December 31, 2019). For those who are still alive, the observation end point is set as the deadline. Crude regression estimates are presented and estimates adjusted for covariates are presented. Confounding factors were selected according to literature and clinical significance, and multivariate models were constructed. We adjusted for the following covariates: age (years), gender, ethnicity, ratio of family income to poverty, smoking status, alcohol use, physical activity, BMI (kg/m²), diabetes, and hypertension.

We used a generalized additive model (GAM) and a restricted cubic spline to model PHQ-9 score and risk of death in our preliminary analysis to visualize its dose-response curve. If a significant nonlinear association is found in the spline model, we further apply the two-piece-wise Cox regression model to characterize this relationship, as described in the previous analysis²⁴. We perform the likelihood ratio test to evaluate the superiority of the two-stage model over the linear model. If the P-value shows that the two-stage model is significantly better (P < 0.05), a statistical threshold effect is considered to exist, as described in the previous analysis^24,25.

To examine the robustness of the results, we conducted sensitivity analyses. Dummy variables were used to indicate missing covariate values, which was performed when continuous variables were missing more than 1% of value. Incorporating PHQ-9 scores into Cox models according to categorical variables (e.g., 0–4, 5–9, 10–14, ≥ 15) to verify trend consistency.

The two-sided alpha level was set at 0.05. All the statistical analyses were performed using the EmpowerStats (www.empowerstats.com, X&Y solutions, Inc. Boston MA) and R software version 3.6.1 (http://www.r-project.org)²⁶.

Results

Baseline characteristics

The study included 36,393 participants with a mean age of 47.9 years (SD = 18.8). During a median follow-up of 89 months, 3,644 (10.01%) deaths occurred. The distribution of depression symptoms was as follows: no depression (75.26%), mild depression (15.95%), moderate depression (5.48%) and severe depression (3.31%).

The total sample size was 36,393, with 33,193 in the no depression group and 3,200 in the depression group. There was no significant difference in mean age at screening between the two groups (P = 0.879). The proportion of females was significantly higher in the depression group compared to the no depression group (P < 0.001).The distribution of racial and Hispanic groups differed significantly between the depression and no depression groups (P < 0.001).The proportion of individuals with high BMI was significantly higher in the depression group (P < 0.001). Depressed individuals had a lower poverty income ratio, with a greater proportion in the poor (38.29% vs. 20.29%) or nearly poor categories (32.03% vs. 26.21%; P < 0.001). Depressed individuals were more likely to be current smokers (36.90% vs. 19.33%) and to have diabetes (24.04% vs. 15.70%) or hypertension (58.50% vs. 50.34%), all P < 0.001. Physical activity levels were lower among the depressed group, with a higher proportion classified as inactive (72.71% vs. 65.24%; P < 0.001) (Table 1).

Table 1.

Baseline characteristics by depression status.

Parameters	No depression (PHQ-9 < 10) (N = 33,193)	Depression (PHQ-9 ≥ 10) (N = 3,200)	P-value
Age (years)	47.87 ± 18.95	47.92 ± 17.20	0.879
Gender, N (%)			< 0.001
Male	16,715 (50.36%)	1181 (36.91%)
Female	16,478 (49.64%)	2019 (63.09%)
Ethnicity, N (%)			< 0.001
Mexican	5388 (16.23%)	510 (15.94%)
Hispanic	3019 (9.10%)	413 (12.91%)
White	13,925 (41.95%)	1324 (41.38%)
Black	7237 (21.80%)	712 (22.25%)
Other	3624 (10.92%)	241 (7.53%)
Education level, N (%)			< 0.001
< 9th Grade	3039 (9.75%)	458 (15.02%)
9-11th grade	4213 (13.51%)	640 (20.98%)
High school graduate	7175 (23.01%)	738 (24.20%)
Some college or AA degree	9250 (29.66%)	907 (29.74%)
College graduate or above	7508 (24.08%)	307 (10.07%)
Marital status, N (%)			< 0.001
Married	16,549 (52.24%)	1077 (34.98%)
Widowed	2434 (7.68%)	300 (9.74%)
Divorced	3236 (10.22%)	538 (17.47%)
Separated	957 (3.02%)	213 (6.92%)
Never married	5960 (18.81%)	647 (21.01%)
Living with partner	2542 (8.02%)	304 (9.87%)
Poverty income ratio, N (%)			< 0.001
Poor	6169 (20.29%)	1108 (38.29%)
Nearly poor	7969 (26.21%)	927 (32.03%)
Middle income	8236 (27.09%)	555 (19.18%)
High income	8029 (26.41%)	304 (10.50%)
BMI (kg/m2)	28.72 ± 6.70	30.27 ± 8.07	< 0.001
Smoking status, N (%)			< 0.001
Never	18,217 (57.04%)	1281 (41.14%)
Former	7549 (23.63%)	684 (21.97%)
Current	6174 (19.33%)	1149 (36.90%)
Alcohol use, drinks/day			< 0.001
0	8682 (26.99%)	815 (26.10%)
1	19,781 (61.49%)	1914 (61.29%)
2	2174 (6.76%)	183 (5.86%)
≥ 3	1531 (4.76%)	211 (6.76%)
Total physical activity			< 0.001
Inactive participants	18,701 (65.24%)	2105 (72.71%)
Active participants	9962 (34.76%)	790 (27.29%)
Diabetes, N (%)			< 0.001
No	27,959 (84.30%)	2430 (75.96%)
Yes	5207 (15.70%)	769 (24.04%)
Hypertension, N (%)			< 0.001
No	16,480 (49.66%)	1328 (41.50%)
Yes	16,708 (50.34%)	1872 (58.50%)

BMI, body mass index; PHQ, patient health questionnaire. Results are presented as Mean ± SD for continuous variables and N (%) for categorical variables. Among the 36,393 participants, education level had 2158 (5.9%) missing values, marital status had 1636 (4.5%) missing values, poverty income ratio had 3096 (8.5%) missing values, BMI had 17,524 (48.2%) missing values, smoking status had 1339 (3.7%) missing values, alcohol use had 1102 (3.0%) missing values, total physical activity had 4835 (13.3%) missing values, diabetes had 28 (0.1%) missing values, and hypertension had 5 (0.01%) missing values.

Association between demographic, clinical, and socioeconomic factors and All-cause mortality

Age was positively associated with mortality, increasing the risk by 9% for each year (HR 1.09, P < 0.001). Females had lower risks than males (HR 0.70, P < 0.001), while racial disparities showed higher mortality in Whites and Blacks compared to Mexicans (HR 2.97 and HR 2.07, respectively, P < 0.001). Higher education levels correlated with lower mortality risk (HR 0.40 for college graduates, P < 0.001). Marital status significantly affected risk, as widowed individuals had much higher mortality (HR 4.52, P < 0.001), while single and partnered individuals had lower risks (HR 0.44 and HR 0.50, P < 0.001). Diabetes and hypertension were linked to increased mortality risk. Additionally, physical inactivity (HR 0.39, P < 0.001), smoking, and heavy alcohol use were associated with higher mortality. Depressive symptoms also correlated with increased risk, particularly severe depression (HR 1.47, P < 0.001) (Table 2).

Table 2.

The unadjusted association between baseline variables and all-cause mortality (n = 36,393).

Exposure	Statistics	HR (95% CI)	P value
Age (years)	47.87 ± 18.80	1.09 (1.09, 1.09)	< 0.001
Gender
Male	17,896 (49.17%)	Reference
Female	18,497 (50.83%)	0.70 (0.65, 0.75)	< 0.001
Ethnicity
Mexican	5898 (16.21%)	Reference
Hispanic	3432 (9.43%)	1.24 (1.03, 1.48)	0.020
White	15,249 (41.90%)	2.97 (2.64, 3.35)	< 0.001
Black	7949 (21.84%)	2.07 (1.81, 2.36)	< 0.001
Other	3865 (10.62%)	1.03 (0.84, 1.25)	0.805
Education
< 9th Grade	3497 (10.21%)	Reference
9-11th grade	4853 (14.18%)	0.82 (0.74, 0.92)	0.005
High school graduate	7913 (23.11%)	0.73 (0.66, 0.81)	< 0.001
Some college or AA degree	10,157 (29.67%)	0.53 (0.47, 0.59)	< 0.001
College graduate or above	7815 (22.83%)	0.40 (0.36, 0.45)	< 0.001
Marital status
Married	17,626 (50.71%)	Reference
Widowed	2734 (7.87%)	4.52 (4.18, 4.90)	< 0.001
Divorced	3774 (10.86%)	1.47 (1.33, 1.62)	< 0.001
Separated	1170 (3.37%)	1.11 (0.92, 1.34)	0.261
Never married	6607 (19.01%)	0.44 (0.39, 0.50)	< 0.001
Living with partner	2846 (8.19%)	0.50 (0.42, 0.59)	< 0.001
Poverty income ratio
Poor	7277 (21.85%)	Reference
Nearly poor	8896 (26.72%)	1.37 (1.25, 1.50)	< 0.001
Middle income	8791 (26.40%)	1.00 (0.91, 1.10)	0.971
High income	8333 (25.03%)	0.55 (0.49, 0.62)	< 0.001
BMI (kg/m2) groups
< 18.5	342 (1.81%)	Reference
≥ 18.5, < 25	5443 (28.85%)	0.67 (0.52, 0.86)	0.017
≥ 25	13,084 (69.34%)	0.70 (0.55, 0.89)	0.036
Diabetes
No	30,389 (83.57%)	Reference
Yes	5976 (16.43%)	3.14 (2.93, 3.36)	< 0.001
Hypertension
No	17,808 (48.94%)	Reference
Yes	18,580 (51.06%)	3.99 (3.69, 4.32)	< 0.001
Total physical activity
Inactive participants	20,806 (65.93%)	Reference
Active participants	10,752 (34.07%)	0.39 (0.35, 0.43)	< 0.001
Smoking status
Never	19,498 (55.62%)	Reference
Former	8233 (23.49%)	2.42 (2.25, 2.61)	< 0.001
Current	7323 (20.89%)	1.48 (1.36, 1.62)	< 0.001
Alcohol use, drinks/day
0	9497 (26.91%)	Reference
1	21,695 (61.47%)	0.83 (0.77, 0.89)	< 0.001
2	2357 (6.68%)	0.64 (0.55, 0.75)	< 0.001
≥ 3	1742 (4.94%)	0.87 (0.74, 1.01)	0.069
Depressive symptoms
No depression	27,389 (75.26%)	Reference
Mild depression	5804 (15.95%)	1.26 (1.15, 1.37)	< 0.001
Moderate depression	1995 (5.48%)	1.39 (1.22, 1.58)	< 0.001
Severe depression	1205 (3.31%)	1.47 (1.25, 1.72)	< 0.001

Data are expressed as Mean ± SD for continuous variables and N (%) for categorical variables. BMI, body mass index; CI, confidence interval; HR, hazard ratio.

Identification of nonlinear relationship

A nonlinear dose–response relationship was observed between PHQ-9 depression scores and all-cause mortality (Fig. 2). Log(relative risk) can be converted to a relative risk by taking antilog. For example, a log(relative risk) of 0 implies the relative risk of 1 (no impact on the probability of death), whereas a log(relative risk) of 1 implies the relative risk of 2.71 (i.e., 2.71-fold increase in the probability of death). Based on the generalized additive model, when the PHQ-9 score was low, the risk of death increased concomitantly with an increase in PHQ-9 score. However, once the PHQ-9 score surpassed a certain threshold, the observed increase in mortality risk plateaued.

Fig. 2.

Dose-response relationships between depression score and the probability of all-cause mortality. A nonlinear association between depression score and all-cause mortality was found in a generalized additive model. The resulting figures show the log(relative risk) in the y-axis and the continuous covariate in the x-axis. The solid line and dashed line represent the estimated values and their corresponding 95% confidence intervals. Adjustment factors included age (years), gender, ethnicity, ratio of family income to poverty, smoking status, alcohol use, physical activity, BMI (kg/m²), diabetes, and hypertension. Abbreviations: BMI, body mass index; RR, relative risk; CI, confidence interval.

Using a threshold effect analysis, a two-piece-wise linear regression model further confirmed the nonlinear association between depression scores and mortality (Table 3). The turning point for depression scores was identified at 7 for both all-cause and cardiovascular disease mortality, with a 95% confidence interval of 6 to 10. Below the threshold, each unit increase in depression score was associated with a significantly higher risk of all-cause (HR 1.06, 95% CI 1.05–1.08, P < 0.001) and cardiovascular mortality (HR 1.07, 95% CI 1.03–1.10, P < 0.001). However, above the threshold, no significant association was observed (all-cause mortality HR 1.01, P = 0.238; cardiovascular mortality HR 0.99, P = 0.675). The log-likelihood ratio test indicated that the two-piece-wise linear model was significantly better than the linear model for both outcomes (P < 0.001 and P = 0.010, respectively) (Fig. 2; Table 3).

Table 3.

Threshold effect analysis of depression score and the probability of all-cause mortality and cardiovascular disease mortality.

Models	All-cause mortality		Cardiovascular disease mortality
	HR (95%CI)	P value	HR (95%CI)	P value
Model I
One line effect	1.04 (1.03, 1.04)	< 0.001	1.03 (1.02, 1.05)	< 0.001
Model II
Turning point (K)	7		7
depression score < K	1.06 (1.05, 1.08)	< 0.001	1.07 (1.03, 1.10)	< 0.001
depression score ≥ K	1.01 (0.99, 1.03)	0.238	0.99 (0.96, 1.03)	0.675
P value for LRT test*		< 0.001		0.010
95% CI for turning point	6, 10		6, 10

Data were presented as HR (95% CI) P value; Model I, linear analysis; Model II, non-linear analysis. Adjusted for age (years), gender, ethnicity, ratio of family income to poverty, smoking status, alcohol use, BMI (kg/m²), physical activity, diabetes, and hypertension. BMI, body mass index; CI, confidence interval; HR, hazard ratio; LRT, logarithm likelihood ratio test. * P < 0.05 indicates that model II is significantly different from Model I.

Kaplan-Meier survival analysis by depression status

The Kaplan-Meier analysis demonstrated that individuals with depression (PHQ-9 ≥ 10) had significantly lower survival probabilities compared to individuals without depression (PHQ-9 < 10) throughout the follow-up period (P < 0.0001) (Fig. 3).

Fig. 3.

This Kaplan-Meier curve illustrates the survival probability over time for groups: individuals with no depression (PHQ-9 score < 10) and individuals with depression (PHQ-9 score ≥ 10). The y-axis represents the survival probability, and the x-axis represents time. The “Number at risk” table at the bottom shows the number of participants in each group at different time points.

The results of the sensitivity analysis were consistent with the core results. Table 4 demonstrates the adjusted associations between depression levels, measured by PHQ-9 scores, and all-cause mortality. In the non-adjusted model, each unit increase in PHQ-9 score was associated with a 3% higher risk of mortality (HR = 1.03, 95%CI 1.02–1.03, P < 0.001); after adjustment for confounders in Adjust I and Adjust II models, the associations remained significant (HR = 1.04, 95%CI:1.03–1.05, P < 0.001; HR = 1.04, 95%CI: 1.03–1.04, P < 0.001). Compared with participants without depressive symptoms, those with mild, moderate, and severe depression consistently showed significantly higher risks of mortality across all models. In the fully adjusted model (Adjust II), mild, moderate, and severe depression were associated with 31% (HR = 1.31, 95%CI: 1.20–1.44, P < 0.001), 47% (HR = 1.47, 95%CI: 1.27–1.69, P < 0.001), and 57% (HR = 1.57, 95%CI: 1.31–1.87, P < 0.001) increased risk of mortality, respectively.

Table 4.

Sensitivity analysis of depression score associated with all-cause mortality.

Exposure	Person-y	No. of Events	Mortality Rate (per 1000 Person-y)	HR (95%CI) P value
				Adjust I	Adjust II
PHQ-9 score	276088.1	3644	13.19	1.04 (1.03, 1.05) < 0.001	1.04 (1.03, 1.04) < 0.001
Depressive symptoms
No	209754.1	2580	12.3	1.0 (Reference)	1.0 (Reference)
Mild	42901.2	659	15.36	1.35 (1.23, 1.48) < 0.001	1.31 (1.20, 1.44) < 0.001
Moderate	14696.5	249	16.94	1.55 (1.35, 1.78) < 0.001	1.47 (1.27, 1.69) < 0.001
Severe	8736.3	156	17.85	1.68 (1.41, 2.01) < 0.001	1.57 (1.31, 1.87) < 0.001

Adjust I model adjust for: age (years), gender, ethnicity, ratio of family income to poverty, smoking status, alcohol use, and BMI (kg/m²).

Adjust II model adjust for: age (years), gender, ethnicity, ratio of family income to poverty, smoking status, alcohol use, BMI (kg/m²), physical activity, diabetes, and hypertension.

BMI, body mass index; CI, confidence interval; HR, hazard ratio.

Discussion

Our study found a non-linear association between depression scores and all-cause mortality, with a threshold effect. This suggests that even mild to moderate depression is associated with a significant increase in mortality risk, while the risk increase plateaus at higher depression scores.

Numerous studies have explored the association between depression and health outcomes using data from the NHANES. A key focus has been the impact of depression, assessed via the PHQ-9, on all-cause and cardiovascular mortality^27,28. These studies consistently demonstrate a dose-response relationship, where the risk of mortality increases with the severity of depressive symptoms^5,28, underscoring the profound link between mental and physical health.Our findings align with and extend previous research, demonstrating a non-linear relationship between depressive symptoms and mortality risk, with a threshold effect at a PHQ-9 score of 7. This indicates that even mild depressive symptoms significantly elevate mortality risk in the general population, while the risk associated with severe depression may plateau, potentially due to competing risk factors such as comorbidities or advanced disease states^29,30. This pattern is consistent with prior studies, which have also observed that mild to moderate depressive symptoms are linked to a higher mortality risk, whereas severe depression may not show a proportional increase in risk, possibly due to the influence of other health conditions^30,31.The heterogeneity in findings across studies may arise from differences in study design, population characteristics, and the specific health outcomes examined^29,32.

The potential mechanisms linking depression to all-cause mortality are multifaceted, may include biological factors such as inflammation and dysregulation of the hypothalamic-pituitary-adrenal (HPA) axis^33,34.Both of them can exacerbate various health conditions, including mental health disorders^35,36. In addition to biological factors, depression often leads to unhealthy lifestyle choices, including poor diet, physical inactivity, smoking, and substance abuse^37,38.These behaviors also increase the likelihood of developing new comorbidities, further elevating mortality risk. Cognitive mechanisms are equally important. Cognitive impairment frequently coexists with depression^39,40. The presence of cognitive deficits may hinder an individual’s ability to seek and adhere to treatment, thereby worsening their overall health status and increasing mortality risk⁴¹.Evidence suggests that effective management of depressive symptoms, whether through pharmacological or psychosocial interventions, can lead to improved health outcomes and reduced mortality rates⁴².

Study limitations

Methodologically, our findings address key limitations of prior studies. The large, nationally retrospective sample (N = 36,393) enhances generalizability over clinic-based cohorts, as evidenced by the longitudinal design and comprehensive data collection methods employed in recent research⁴³. This approach minimizes biases often associated with smaller, clinic-based samples, thereby providing a more accurate reflection of the population. Additionally, the retrospective design of our study significantly reduces recall bias inherent in cross-sectional analyses, which is a well-documented limitation in the literature.

Conclusions

Depression scores exhibit a non-linear association with all-cause mortality, with a significant increase in risk at lower scores and a plateau effect at higher scores. These findings underscore the importance of early identification and intervention for depression, even at lower severity levels, to reduce mortality risk. Future research should explore the mechanisms underlying this relationship and evaluate the effectiveness of interventions targeting mild to moderate depression in reducing mortality.

Author contributions

Y.H. cleaned the data. X.C.(Xinglin Chen) performed statistical analysis.X.C.(Xiaolan Cai) conceived and designed the research. Y.H. and X.C.(Xinglin Chen) drafted the manuscript. X.C.(Xiaolan Cai) made critical revision of the manuscript for key intellectual content.

Data availability

The data used in this study are publicly available from the NHANES database (https://www.cdc.gov/nchs/nhanes/index.htm) and the National Death Index (https://www.cdc.gov/nchs/data-linkage/mortality-public.htm).

Declarations

Competing interests

The authors declare no competing interests.

Ethics statement

The NHANES study was approved by the National Center for Health Statistics (NCHS) Research Ethics Review Board, and all participants provided written informed consent.