Trends in multimorbidity among US adults and the correlation between physical activity and mortality: Prospective cohort study from 1999 to 2018

Abstract

Evidence regarding the mortality risk associated with physical activity (PA) and multimorbidity is limited. This study investigates trends in the prevalence of multimorbidity in the population and the mortality risk associated with PA and multimorbidity. The association between PA and overall survival was analyzed in 39,940 subjects from the National Health and Nutrition Examination Survey cohorts. Among the 39,940 participants (weighted average age: 46.84; 50.86% female), the age-standardized prevalence of multimorbidity rose significantly, from 15.87% (95% confidence interval [CI] = 14.86%–16.87%) in 1999 to 2000 to 19.96% (95% CI = 17.61%–22.32%) in 2017 to 2018. The population’s multimorbidity detection rate was 25.22%. The all-cause mortality risks for multimorbidity were: dual-multimorbidity (hazard ratio [HR] = 1.97, 95% CI = 1.76–2.20), triple-multimorbidity (HR = 2.69, 95% CI = 2.35–3.08), quadruple-multimorbidity (HR = 3.22, 95% CI = 2.59–3.99). “Diabetes mellitus + cardiovascular disease + respiratory diseases” had the highest mortality risk (HR = 4.37, 95% CI = 2.67–7.16). PA reduced all-cause and cause-specific mortality, with moderate-intensity PA lowering all-cause mortality by 22% compared to low-intensity PA, and high-intensity PA by 6% compared to moderate-intensity PA. Multimorbidity led to a 5.64 years life expectancy loss, while PA could recover up to 2.31 years. PA potentially recouped 2.96 and 2.19 years of life lost due to comorbidity in females and males, respectively. According to National Health and Nutrition Examination Survey data from 1999 to 2000 to 2017 to 2018, multimorbidity prevalence increased. PA reduces all-cause mortality by 32%. Greater PA intensity correlates with more significant mortality reduction. Multimorbidity cuts life expectancy by 5.64 years, recoverable by PA (2.31 years). Females regain 2.96 years; males, 2.19 years.

1. Introduction

Global noncommunicable disease (NCD) prevalence and mortality surged from 2000 to 2019,^[1,2] with NCD deaths rising by over a third, from 31 to 41 million, representing 74% of global fatalities. NCDs accounted for 47% of the global burden in 2000, escalating to 63% by 2019, projected to hit 86% by 2048.^[1,2] Multimorbidity individuals face heightened mortality risks, impacting clinical management, reducing quality of life, and increasing socioeconomic burdens.^[3,4] Physical activity (PA) mitigates NCD risks and multimorbidity’s adverse effects, playing a critical role in prevention and management.^[2,5,6] This National Health and Nutrition Examination Survey (NHANES)-based study analyzes US adult chronic disease prevalence, assesses multimorbidity’s mortality risk, identifies high-risk combinations, and explores PA’s role in mortality risk reduction and life expectancy recovery.

2. Materials and methods

2.1. Data collection

NHANES, launched in 1999, is a national cross-sectional survey accessible via the Centers for Disease Control and Prevention website. Approved by the National Center for Health Statistics (NCHS) Ethics Review Committee, it involves noninstitutionalized civilians. Participants, aged ≥20, provided written consent. Data from 10 cycles (1999–2018) were analyzed, excluding individuals <20, pregnant, lost to follow-up, with <12-month follow-up, MET of 0, or missing covariates. See Figure S1, Supplemental Digital Content 1 for a detailed enrollment flowchart, including exclusion of participants with missing covariates. The study comprised 39,940 participants and required no further ethical review or consent due to NCHS approval (45 CFR §46.102(f)).

Age, gender, race/ethnicity, education, marital status, poverty level, smoking, and alcohol use data were gathered via household interviews. Trained interviewers collected PA information using a global PA questionnaire. Weekly PA metabolic equivalent (MET) was computed by multiplying moderate or vigorous PA minutes and frequency by NHANES-recommended MET values. Weight, height, and blood pressure were measured at examination centers following standard procedures. Body mass index (BMI) was derived from weight (kg) divided by height squared (m). The mean of all blood pressure readings determined systolic blood pressure (SBP) and diastolic blood pressure (DBP).

2.2. Definition of diseases

Hypertension (HT; presence of ≥1 criterion): SBP ≥ 140 mm Hg or DBP ≥ 90 mm Hg (mean of 3 readings); current use of antihypertensive medication; self-reported physician diagnosis; or known HT. Diabetes mellitus (DM; presence of ≥1 criterion): self-reported physician diagnosis; hemoglobin A1c ≥ 6.5%; fasting glucose ≥ 7.0 mmol/L; random glucose ≥ 11.1 mmol/L; oral glucose tolerance test glucose ≥ 11.1 mmol/L; current use of diabetes medication/insulin; or impaired fasting glucose/tolerance. Cardiovascular disease (CVD; self-reported history of ≥1 condition via questionnaire): congestive heart failure, coronary heart disease, angina, or heart attack. Respiratory diseases (RES; defined as chronic obstructive pulmonary disease or asthma): chronic obstructive pulmonary disease is bronchodilator use with FEV1/FVC < 0.7, doctor diagnosis, age ≥40 with smoking history, or bronchitis history and medication use. Asthma is diagnosed by a doctor, bronchodilator use, age <40 with no smoking, bronchitis, or emphysema history, and medication use. Individuals meeting ≥2 condition definitions were classified into multimorbidity groups; others comprised the reference group (disease-free for all 4 conditions).

NHANES mortality files provide data up to December 31, 2019, using International Classification of Diseases, Injuries, and Causes of Death-10th Revision codes. We selected cause-specific mortality: heart diseases (I00–I09, I11, I13, I20–I51), lower RES (J40–J47), and diabetes (E10–E14) based on NHANES guidelines.

2.3. Outcomes

Primary outcomes: prevalence of multimorbidity among US adults, effects of multimorbidity patterns on overall and cause-specific mortality risks, and PA’s impact on life expectancy. Secondary outcomes: prevalence of 4 chronic diseases and multimorbidity across different subgroups.

2.4. Statistical analysis

Using the complete NHANES sample (1999–2018, n = 39,940), we assessed trends in chronic disease and multimorbidity prevalence. We reported P values for overall trends. Chi-square tests/analysis of variance analyzed baseline characteristics, with categorical variables as percentages (95% confidence interval [CI]) and continuous data as mean ± standard error. Subgroup analyses considered age, sex, race/ethnicity, education, BMI, poverty impact ratio (PIR), and trends. Cox regression, adjusting for demographics, assessed hazard ratio (HR) and 95% CI. A decision tree model evaluated multimorbidity’s impact on mortality risk. PA, analyzed using restricted cubic splines (RCS), categorized participants into 3 MET groups. These cohort-derived activity thresholds, optimized for intra-cohort risk stratification, are not absolute metrics. We explored PA’s impact on mortality risk in multimorbid individuals, adjusting for confounding factors. NCHS demographic data (1999–2018) created life expectancy tables spanning ages 20 to 100 for each year.^[7,8] R 4.2.2 software (R Foundation for Statistical Computing), with NHANES sample weights, conducted analyses.^[9,10] P < .05 was considered significant. Complete-case analysis was employed, excluding all individuals with missing data on exposure, outcome, or covariates (final n = 39,940).

3. Results

3.1. Population characteristics

In 10 NHANES survey cycles, 39,940 participants provided data, representing 168,017,674 US residents aged 20 and older. Female participants comprised 49.14%, while males were 50.86%. Racially, 70.41% were non-Hispanic White (White), 7.67% Mexican American (Mexican), 10.57% non-Hispanic Black (Black), and 11.35% other races (other Hispanic and other race, including multi-racial). Median follow-up time was 113 months (max: 250 months, average: 120.79 ± 1.18 months). Multimorbidity groups showed significant differences in survey cycle, age, race, education, PIR, BMI, smoking, drinking, activity level, survival status, follow-up time, and MET (P < .001; Table 1).

Table 1.

Study participants’ baseline characteristics under multimorbidity patterns, NHANES, 1999 to 2018.^*

Variable	Total (n = 39,940)†	Healthy population‡ (n = 16,416)†	Single-disease (n = 13,451)†	Dual-multimorbidity (n = 7387)†	Triple-multimorbidity (n = 2290)†	Quadruple-multimorbidity (n = 396)†	P value
Years							<.0001
1999–2000	7.36 (6.04, 8.67)	7.85 (6.68, 9.02)	7.77 (6.16, 9.38)	5.58 (4.42, 6.73)	5.44 (3.78, 7.09)	6.31 (3.68, 8.94)
2001–2002	9.72 (8.88, 10.56)	10.88 (9.83, 11.92)	9.57 (8.63, 10.51)	7.67 (6.63, 8.72)	6.51 (4.94, 8.08)	5.55 (1.49, 9.61)
2003–2004	9.63 (8.29, 10.97)	10.24 (8.97, 11.50)	9.81 (8.57, 11.04)	7.79 (6.33, 9.25)	8.49 (5.15, 11.83)	8.57 (4.82, 12.31)
2005–2006	10.40 (9.10, 11.71)	10.61 (9.22, 12.00)	10.38 (8.89, 11.87)	10.13 (8.56, 11.70)	10.04 (7.45, 12.64)	6.57 (4.11, 9.04)
2007–2008	10.28 (9.04, 11.53)	9.83 (8.73, 10.93)	10.36 (8.93, 11.79)	11.37 (9.59, 13.15)	10.13 (8.09, 12.16)	13.94 (8.67, 19.21)
2009–2010	10.29 (9.08, 11.50)	10.40 (9.27, 11.54)	9.65 (8.42, 10.88)	10.95 (9.28, 12.63)	11.76 (9.50, 14.02)	9.89 (5.86, 13.93)
2011–2012	10.61 (9.24, 11.98)	10.15 (8.90, 11.41)	10.55 (8.93, 12.17)	11.32 (9.62, 13.02)	13.29 (10.08, 16.49)	11.47 (5.85, 17.10)
2013–2014	11.31 (10.01, 12.62)	10.50 (9.08, 11.92)	11.58 (10.18, 12.99)	12.69 (11.58, 13.80)	12.69 (10.59, 14.78)	13.96 (7.23, 20.69)
2015–2016	11.13 (9.81, 12.45)	10.37 (9.06, 11.68)	10.96 (9.42, 12.50)	13.17 (11.47, 14.86)	12.51 (10.28, 14.75)	15.86 (9.07, 22.65)
2017–2018	9.26 (8.64, 9.88)	9.18 (8.16, 10.21)	9.38 (8.46, 10.30)	9.34 (7.87, 10.80)	9.14 (6.92, 11.37)	7.87 (2.69, 13.05)
Age (yr)							<.0001
20–39	36.84 (35.32, 38.37)	53.03 (51.58, 54.47)	30.31 (29.02, 31.59)	12.70 (11.62, 13.77)	5.43 (4.11, 6.76)	1.02 (0.12, 1.92)
40–49	20.42 (19.42, 21.42)	22.71 (21.72, 23.70)	20.70 (19.72, 21.68)	16.31 (14.97, 17.66)	9.74 (8.08, 11.39)	12.24 (8.19, 16.29)
50–59	18.62 (17.54, 19.69)	14.83 (13.98, 15.68)	21.03 (19.96, 22.11)	24.03 (22.68, 25.37)	21.34 (19.03, 23.66)	19.62 (13.65, 25.58)
60–69	13.06 (12.26, 13.86)	6.25 (5.62, 6.87)	15.32 (14.49, 16.15)	22.66 (21.31, 24.02)	31.03 (28.35, 33.72)	33.14 (26.76, 39.53)
70–79	7.62 (7.13, 8.11)	2.47 (2.19, 2.75)	8.57 (7.92, 9.22)	16.26 (15.33, 17.19)	21.61 (19.26, 23.96)	26.21 (21.08, 31.34)
≥80	3.45 (3.16, 3.74)	0.72 (0.61, 0.82)	4.07 (3.69, 4.45)	8.04 (7.33, 8.75)	10.84 (9.24, 12.45)	7.77 (5.32, 10.23)
Sex							.56
Female	50.86 (48.83, 52.89)	51.32 (50.49, 52.15)	50.71 (49.67, 51.75)	50.18 (48.59, 51.77)	50.06 (47.22, 52.90)	47.42 (40.67, 54.17)
Male	49.14 (47.22, 51.07)	48.68 (47.85, 49.51)	49.29 (48.25, 50.33)	49.82 (48.23, 51.41)	49.94 (47.10, 52.78)	52.58 (45.83, 59.33)
Race/ethnicity							<.0001
White	70.41 (66.00, 74.82)	68.98 (66.97, 71.00)	71.59 (69.46, 73.72)	70.96 (68.60, 73.33)	74.49 (71.80, 77.18)	70.59 (64.61, 76.57)
Mexican	7.67 (6.72, 8.62)	9.29 (8.11, 10.47)	6.84 (5.85, 7.83)	5.91 (4.79, 7.03)	3.73 (2.72, 4.73)	3.14 (1.13, 5.15)
Black	10.57 (9.66, 11.48)	9.31 (8.33, 10.30)	10.96 (9.72, 12.19)	12.71 (11.24, 14.17)	12.75 (10.80, 14.69)	15.26 (11.78, 18.74)
Other	11.35 (10.44, 12.26)	12.41 (11.25, 13.57)	10.61 (9.60, 11.63)	10.42 (9.23, 11.61)	9.04 (7.52, 10.55)	11.00 (6.64, 15.37)
Education							<.0001
Less than high school	5.13 (4.74, 5.53)	3.98 (3.57, 4.40)	4.93 (4.43, 5.43)	7.45 (6.65, 8.24)	9.85 (8.42, 11.28)	10.60 (7.07, 14.13)
High school or equivalent	34.66 (32.72, 36.60)	32.16 (30.67, 33.65)	35.15 (33.56, 36.74)	38.89 (37.18, 40.60)	40.77 (38.08, 43.46)	45.81 (38.40, 53.23)
College or above	60.21 (57.53, 62.89)	63.86 (62.19, 65.52)	59.92 (58.24, 61.61)	53.66 (51.67, 55.66)	49.38 (46.43, 52.33)	43.59 (36.14, 51.04)
Marital							.07
Alone	35.64 (34.41, 36.87)	35.76 (34.38, 37.15)	34.90 (33.60, 36.21)	35.80 (34.27, 37.34)	38.31 (35.72, 40.91)	41.87 (35.13, 48.60)
Married or partner	64.36 (61.25, 67.47)	64.24 (62.85, 65.62)	65.10 (63.79, 66.40)	64.20 (62.66, 65.73)	61.69 (59.09, 64.28)	58.13 (51.40, 64.87)
PIR							<.0001
≤1.0	13.54 (12.70, 14.38)	13.36 (12.42, 14.31)	12.75 (11.85, 13.66)	14.20 (12.91, 15.48)	17.51 (15.65, 19.37)	22.96 (18.06, 27.85)
1.0–3.0	35.63 (33.86, 37.41)	33.74 (32.36, 35.12)	35.65 (34.31, 37.00)	38.66 (37.03, 40.29)	43.16 (40.32, 46.00)	46.18 (39.41, 52.94)
>3.0	50.83 (48.20, 53.45)	52.90 (51.11, 54.69)	51.59 (49.80, 53.39)	47.14 (45.06, 49.22)	39.34 (36.34, 42.34)	30.87 (24.04, 37.70)
BMI							<.0001
18.5–25	29.38 (28.03, 30.73)	38.75 (37.51, 39.98)	25.17 (24.15, 26.19)	15.92 (14.76, 17.09)	12.59 (10.95, 14.23)	8.96 (4.99, 12.94)
<18.5	1.57 (1.40, 1.74)	2.06 (1.78, 2.34)	1.31 (1.05, 1.56)	0.97 (0.65, 1.28)	0.67 (0.17, 1.17)	0.43 (-0.17, 1.03)
25–29.9	33.17 (31.72, 34.63)	33.71 (32.63, 34.79)	34.67 (33.55, 35.79)	30.38 (28.93, 31.82)	27.73 (25.56, 29.89)	22.17 (16.89, 27.46)
≥30	35.87 (34.22, 37.53)	25.48 (24.48, 26.48)	38.85 (37.59, 40.11)	52.73 (51.08, 54.39)	59.01 (56.33, 61.69)	68.43 (61.72, 75.14)
Smoke							<.0001
Never	53.43 (51.43, 55.44)	58.18 (56.88, 59.49)	51.80 (50.40, 53.21)	47.93 (46.28, 49.59)	39.63 (36.84, 42.41)	24.46 (19.51, 29.42)
Former	24.94 (23.52, 26.35)	19.07 (18.00, 20.15)	26.49 (25.50, 27.48)	33.95 (32.33, 35.57)	39.26 (36.43, 42.09)	51.62 (45.53, 57.70)
Now	21.63 (20.46, 22.80)	22.74 (21.66, 23.82)	21.70 (20.59, 22.82)	18.11 (16.85, 19.37)	21.11 (18.89, 23.33)	23.92 (18.35, 29.50)
Drinker							<.0001
Never	10.91 (9.95, 11.88)	10.21 (9.06, 11.36)	10.80 (9.85, 11.75)	12.84 (11.77, 13.92)	12.66 (11.07, 14.26)	9.70 (6.89, 12.51)
Former	14.20 (13.18, 15.23)	10.09 (9.35, 10.84)	14.57 (13.56, 15.58)	20.75 (19.18, 22.31)	27.95 (25.24, 30.65)	38.61 (31.42, 45.81)
Mild	36.12 (34.41, 37.83)	35.15 (33.80, 36.50)	36.98 (35.57, 38.40)	37.07 (35.07, 39.08)	36.88 (34.26, 39.49)	34.22 (26.45, 41.99)
Moderate	17.28 (16.38, 18.17)	19.37 (18.55, 20.20)	17.18 (16.22, 18.13)	13.52 (12.36, 14.68)	10.88 (9.03, 12.73)	5.94 (2.18, 9.69)
Heavy	21.48 (20.44, 22.52)	25.17 (23.97, 26.37)	20.47 (19.48, 21.46)	15.82 (14.48, 17.16)	11.63 (9.67, 13.60)	11.53 (6.11, 16.95)
Activity							<.0001
No	21.00 (19.92, 22.08)	15.85 (15.03, 16.68)	20.77 (19.68, 21.86)	31.01 (29.27, 32.75)	38.18 (35.43, 40.93)	45.03 (37.92, 52.14)
Yes	79.00 (75.84, 82.17)	84.15 (83.32, 84.97)	79.23 (78.14, 80.32)	68.99 (67.25, 70.73)	61.82 (59.07, 64.57)	54.97 (47.86, 62.08)
Survival status							<.0001
Alive	89.66 (86.19, 93.13)	96.02 (95.68, 96.35)	89.18 (88.49, 89.86)	79.29 (78.03, 80.55)	66.84 (64.06, 69.62)	62.84 (56.46, 69.22)
Deceased	10.34 (9.67, 11.01)	3.98 (3.65, 4.32)	10.82 (10.14, 11.51)	20.71 (19.45, 21.97)	33.16 (30.38, 35.94)	37.16 (30.78, 43.54)
lg (PA MET)	4.12 ± 0.06	4.87 ± 0.07	4.14 ± 0.08	2.68 ± 0.13	1.59 ± 0.20	0.72 ± 0.53	<.0001
Follow-up time	120.79 ± 1.18	128.87 ± 1.38	120.49 ± 1.45	105.76 ± 1.36	95.58 ± 2.16	87.99 ± 3.85	<.0001

Black = non-Hispanic Black, BMI = body mass index, lg = base-10 logarithm, MET = metabolic equivalent, Mexican = Mexican American, Other = other Hispanic and other race – including multi-racial, PA = physical activity, PIR = poverty impact ratio, White = non-Hispanic White.

^*Statistics are reported for the sample over 10 waves of data that each span 2 years (1999–2018).

^†Proportions were weighted using National Health and Nutrition Examination Survey sample weights to be nationally representative.

^‡Represents a relatively healthy population in this study.

3.2. The prevalence of 4 chronic diseases

Over 20 years, HT, DM, CVD, and RES were reported in 17,004, 9717, 3431, and 6527 individuals, respectively. HT and CVD showed the least fluctuation, with HT consistently high and CVD relatively low. DM and RES had more significant fluctuations, with DM growing fastest (Fig. 1 and Table S1, Supplemental Digital Content 15). Subgroup analysis found all 4 diseases increased with age. Men had a higher incidence of CVD, DM, and HT, while RES favored women. Incidence varied by race; Black individuals had higher CVD, RES, and HT, while Mexican Americans had more DM, the opposite in RES and HT. Higher education correlated with more RES and CVD but less DM and HT. Lower education showed the opposite. Higher BMI associated with greater risk, especially BMI ≥ 30. Diseases had lower rates in higher PIR (Figs. S2–S5, Supplemental Digital Content 2 and Table S2, Supplemental Digital Content 16).

Figure 1.

Prevalence of 4 chronic diseases and multimorbidity among US adults, 1999 to 2018. (A) Four chronic diseases. (B) Multimorbidity. CI = confidence interval, CVD = cardiovascular disease, DM = diabetes mellitus, HT = hypertension, RES = respiratory diseases.

3.3. Multimorbidity population situation

Eleven multimorbidity patterns were identified, with 18.50% (7387 individuals) representing dual-multimorbidity, 5.73% (2290 individuals) triple-multimorbidity, and 0.99% (396 individuals) quadruple-multimorbidity (Fig. 1 and Table S3, Supplemental Digital Content 17). The most common combinations were “HT + DM” in dual-multimorbidity (10.43%, 4166 individuals), “HT + DM + CVD” in triple-multimorbidity (2.53%, 1009 individuals), and “HT + DM + CVD + RES” in quadruple-multimorbidity (0.99%, 396 individuals; Table S3, Supplemental Digital Content 17). Prevalence increased with age, while females had lower rates than males. Black individuals had higher rates, while White individuals had lower rates. Those with less than a high school education were more prone. Individuals with higher BMI or lower income had higher likelihoods (Fig. S6, Supplemental Digital Content 6 and Table S4, Supplemental Digital Content 18). The study shows that PA individuals have a lower incidence of multimorbidity compared to those who are not PA, whether considering the 4 chronic diseases or multimorbidity patterns (Fig. S7, Supplemental Digital Content 7 and Table S5, Supplemental Digital Content 19).

3.4. Association analysis between 4 chronic diseases and mortality

Further multivariate Cox regression analysis revealed significant associations between all 4 chronic diseases and all-cause mortality, even after adjusting for other factors. CVD posed the highest risk, followed by DM (Table S6, Supplemental Digital Content 20). In addition, Kaplan–Meier survival rates for all-cause mortality showed significant differences (P < .001; Fig. S8, Supplemental Digital Content 8). In a multivariate Cox regression analysis considering only deaths related to CVD/DM/HT/RES, after adjusting for various factors, CVD, DM, and RES were significantly associated with the risk of death, while HT showed no significant association. DM had the highest risk association, followed by RES. Kaplan–Meier survival rates indicated significant differences for CVD and RES (P < .001), while DM (P = .16) and HT (P = .788) showed nonsignificant differences (Fig. S8, Supplemental Digital Content 8).

3.5. Association analysis between multimorbidity patterns and mortality

The Cox regression analysis for all-cause mortality associated with multimorbidity reveals a consistent risk pattern: quadruple-multimorbidity (HR = 3.22, 95% CI = 2.59–3.99) > triple-multimorbidity (HR = 2.69, 95% CI = 2.35–3.08) > dual-multimorbidity (HR = 1.97, 95% CI = 1.76–2.20) > single-disease (HR = 1.39, 95% CI = 1.25–1.53) compared to participants who had none of the 4 target chronic conditions (CVD, HT, DM, and RES; Table S7, Supplemental Digital Content 21). This pattern persists when considering cause-specific mortality. Kaplan–Meier survival curves show significant differences (P < .001), reinforcing these findings (Fig. S9, Supplemental Digital Content 9).

Regarding multimorbidity patterns, “DM + CVD” in dual-multimorbidity poses the highest risk of all-cause mortality (HR = 2.68, 95% CI = 2.09–3.43), with a 268% increase in risk compared to participants who had none of the 4 target chronic conditions (CVD, HT, DM, and RES). In triple-multimorbidity, “DM + CVD + RES” presents a significantly higher risk than other patterns (HR = 4.37, 95% CI = 2.67–7.16). Within quadruple-multimorbidity, “DM + CVD + RES + HT” carries the highest risk of all-cause mortality (HR = 3.28, 95% CI = 2.64–4.06; Table S7, Supplemental Digital Content 21). While all-cause mortality risk is highest for “DM + CVD” multimorbidity pattern, when considering cause-specific mortality, cox regression analysis indicates that in dual-multimorbidity,“HT + CVD” poses the highest risk of all-cause mortality (HR = 1.65, 95% CI = 1.30–2.10). In triple-multimorbidity,“DM + CVD + RES” presents a significantly higher risk than other patterns (HR = 3.09, 95% CI = 1.93–4.95). Within quadruple-multimorbidity, “DM + CVD + RES + HT” carries the highest risk of all-cause mortality (HR = 2.38, 95% CI = 1.76–3.22; Table S7, Supplemental Digital Content 21). Decision tree models further support that a higher number of diseases correlates with decreased survival curves and lower survival rates. CVD emerges as the root node, indicating its significant impact on survival outcomes, with all multimorbidity patterns related to CVD showing decreased survival rates (Fig. S10, Supplemental Digital Content 10).

3.6. Association analysis between multimorbidity patterns mediated by PA and mortality

Compared to those without PA, individuals engaging in PA demonstrate a significantly reduced risk of mortality, encompassing all-cause mortality and cause-specific mortality (Table S8, Supplemental Digital Content 22 and Fig. S11, Supplemental Digital Content 11). PA is linked with a 32% decrease in the all-cause mortality risk.

Utilizing RCS in Figure S12, Supplemental Digital Content 12, we visualized the relationship between MET from PA and all-cause mortality in the multimorbidity population, revealing a sigmoidal correlation (nonlinear P < .001). The population was categorized into 3 groups for further analysis: MET ≤ 0.58 (low-intensity physical activity [LPA]), 0.58 < MET ≤ 2.42 (moderate-intensity physical activity [MPA]), and MET > 2.42 (high-intensity physical activity [UPA]).

In comparison to LPA, both UPA and MPA significantly reduce the all-cause mortality risk, with a more pronounced reduction observed with higher PA intensity (Table 2 and Fig. 2). Adjusting for age, sex, education, race/ethnicity, marital status, smoking, drinking, BMI, and PIR, UPA demonstrates a 6% additional relative reduction in overall mortality risk compared to MPA and a 28% additional relative reduction compared to LPA (Table 2). While there is no significant statistical difference in reducing mortality risk for UPA concerning cause-specific mortality, further competing risk model analysis reveals that UPA significantly lowers the incidence rate of mortality risk with prolonged follow-up, and the greater the PA intensity, the lower the relative incidence rate of mortality risk (P < .001; Fig. S13, Supplemental Digital Content 13).

Table 2.

Mortality risk profile under different levels of physical activity intensity.^*

Group	Crude model†		Model 1‡		Model 2§		Model 3‖
	HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)	P
All-cause mortality
MET ≤ 0.58	Ref		Ref		Ref		Ref
0.58 < MET ≤ 2.42	0.58 (0.53–0.63)	<.0001	0.72 (0.67–0.78)	<.0001	0.76 (0.71–0.83)	<.0001	0.78 (0.72–0.84)	<.0001
MET > 2.42	0.50 (0.45–0.56)	<.0001	0.68 (0.62–0.75)	<.0001	0.72 (0.65–0.79)	<.0001	0.72 (0.65–0.79)	<.0001
P for trend		<.0001		<.0001		<.0001		<.0001
Cause-specific mortality¶
MET ≤ 0.58	Ref		Ref		Ref		Ref
0.58 < MET ≤ 2.42	0.84 (0.72–0.97)	.02	0.85 (0.75–0.97)	.01	0.86 (0.76–0.97)	.01	0.87 (0.77–0.99)	.04
MET > 2.42	0.87 (0.76–1.00)	.05	0.88 (0.77–1.00)	.06	0.90 (0.78–1.04)	.17	0.90 (0.78–1.04)	.16
P for trend		.003		.003		.01		.02

BMI = body mass index, CI = confidence interval, HR = hazard ratio, MET = metabolic equivalent, PIR = poverty impact ratio.

^*The values for HR (95% CI) are weighted using National Health and Nutrition Examination Survey sample weights to be nationally representative.

^†Adjusted for none.

^‡Adjusted for age, sex, education, race/ethnicity, and marital.

^§Adjusted for age, sex, education, race/ethnicity, marital, smoke, drinker, and BMI.

^‖Adjusted for age, sex, education, race/ethnicity, marital, smoke, drinker, BMI, and PIR.

^¶Cardiovascular disease, diabetes mellitus, hypertension, and respiratory diseases mortality.

Figure 2.

Kaplan–Meier survival curves for multimorbidity patterns mortality at different levels of physical activity intensity. MET = metabolic equivalent.

Further analysis demonstrates that under various PA intensities, the all-cause mortality risk due to multimorbidity significantly decreases. Irrespective of the number of multimorbidities, higher PA intensity is associated with a greater relative reduction in all-cause mortality risk (Table S9, Supplemental Digital Content 23 and Fig. S14, Supplemental Digital Content 14). Adjusting for all confounding factors, although MPA can relatively reduce all-cause mortality risk compared to LPA in quadruple-multimorbidity, UPA increases all-cause mortality risk by 11% compared to MPA. Nonetheless, UPA still shows a relative reduction of 53% in all-cause mortality risk compared to MPA.

3.7. Association analysis of PA with the life expectancy loss caused by multimorbidity

Compared to the participants who had none of the 4 target chronic conditions (CVD, HT, DM, and RES), the diseased population (TDP, including single-disease, dual-multimorbidity, triple-multimorbidity, and quadruple-multimorbidity) faced an average life expectancy reduction of 5.64 years (P < .01). The single-disease population (TSDP) and the multimorbidity population (TMP) experienced reductions of 5.79 and 5.53 years, respectively (both P < .01), with females being more vulnerable (Fig. 3A). PA, encompassing LPA, MPA, and UPA interventions, significantly recovered life expectancy in key multimorbidity subgroups: 0.72 years (P < .01) in TDP, 2.08 years (P < .01) in TSDP, and −0.12 years (P = .688) in TMP, with consistently greater recovery observed in females (Fig. 3A). Overall, PA increases life expectancy by 1.49 years compared to non-PA (P < .01). In males, PA boosts life expectancy by 1.10 years (P < .01), while in females, it increases by 2.26 years (P < .01).

Figure 3.

The impact of PA on life expectancy. Error bars represent standard errors. (A) The effect of PA on life expectancy loss under different multimorbidity patterns. (B) The effect of different PA intensities on life expectancy in the overall population. (C) The effect of different PA intensities on life expectancy in the male population. (D) The effect of different PA intensities on life expectancy in the female population. ^*P < .05, ^**P < .01, and ^#P > .05. MPA = moderate-intensity physical activity, PA = physical activity, UPA = high-intensity physical activity. Healthy population: represents a relatively healthy population in this study; all diseased population: including single-disease, dual-multimorbidity, triple-multimorbidity, and quadruple-multimorbidity; multimorbidity: including dual-multimorbidity, triple-multimorbidity, and quadruple-multimorbidity.

MPA and UPA increase life expectancy by 1.08 years (P < .01) and 3.46 years (P < .01), respectively, compared to LPA in the overall population. In participants who had none of the 4 target chronic conditions (CVD, HT, DM, and RES), MPA and UPA raise life expectancy by 1.95 years (P = .07) and 5.80 years (P < .01), respectively. Among TDP, MPA and UPA recover 0.59 years (P = .07) and 2.31 years (P < .01) of lost life. In TSDP, MPA and UPA reclaim 1.22 years (P < .05) and 3.41 years (P < .01) of lost life, respectively. However, in TMP, MPA and UPA recover 0.18 years (P = .64) and 1.50 years (P < .01) of lost life (Fig. 3B).

Gender subgroup analysis reveals that UPA consistently increases life expectancy or recovers more lost years compared to MPA, with females showing more responsiveness (Fig. 3C, D). In males, UPA boosts life expectancy by 3.19 years (P < .01) and 5.12 years (P < .01) in the overall population and in participants who had none of the 4 target chronic conditions (CVD, HT, DM, and RES), and recovers 2.19 years (P < .01) of lost life in TDP (Fig. 3C). In females, PA shows a dose-dependent enhancement in life expectancy (Fig. 3D). Both MPA and UPA recover 1.81 years (P < .01) and 2.96 years (P < .01) of lost life in TDP (Fig. 3C).

4. Discussion

This study provides new insights into multimorbidity trends and the protective effects of PA on mortality risk and life expectancy among US adults, utilizing nationally representative NHANES data over 2 decades (1999–2018).

4.1. Key findings and prevalence trends

Our analysis confirms a significant and concerning rise in multimorbidity prevalence among US adults between 1999 and 2000 and 2017 and 2018. Nearly one-quarter (25.22%) of the sampled US adults were identified as having multimorbidity (≥2 of the 4 chronic conditions studied). This upward trend aligns with growing evidence of an increasing chronic disease burden in the United States.^[11] Similar to prior work,^[12–15] we observed substantial disparities in multimorbidity burden across sociodemographic groups, with higher prevalence and severity concentrated among males, Black individuals, those with lower educational attainment, higher BMI, and lower socioeconomic status. This granular detail on disparities within a representative US sample over this time period adds valuable context to understanding the uneven burden of disease.

Our findings regarding the individual prevalence of CVD, HT, and RES are consistent with established estimates.^[12–15] However, our observed DM prevalence differs from some reports.^[16] This variation could be explained, in part, by our expanded definition of DM within this multimorbidity framework. The dominance of CVD and HT combinations within multimorbidity patterns, consistent with global trends emphasizing their high prevalence,^[13,17–19] is particularly crucial given their established strong link to increased CVD risk.^[20–23] Indeed, we found that multimorbidity patterns involving CVD conveyed the highest mortality risk, reinforcing the critical nature of these specific combinations.

4.2. PA, mortality risk, and life expectancy: mechanisms and comparisons

Our results robustly demonstrate a strong protective association between PA and survival. After extensive multivariable adjustment, physically active individuals experienced a 32% lower risk of all-cause mortality compared to inactive counterparts. More importantly, specifically within the multimorbidity subpopulation, engaging in MPA or UPA reduced all-cause mortality risk by 22% and 26%, respectively. Critically, we observed a dose–response relationship, with higher PA intensity associated with significantly greater mortality risk reduction and recovery of life expectancy lost due to multimorbidity. This graded effect is highly consistent with the established biological benefits of PA demonstrated in prior studies^[24] and aligns with the underlying physiological mechanisms: PA improves insulin sensitivity,^[25] enhances endothelial function,^[26] reduces HT risk,^[27] and counters the deleterious effects of sedentary behavior, particularly harmful when combined with high BMI in relation to CVD mortality.^[28] Furthermore, our findings regarding PA’s association with reduced RES mortality risk add to the growing evidence base supporting PA for RES management.^[29] Our quantification of life expectancy impacts is a notable contribution: multimorbidity was associated with a substantial loss of 5.64 years; participation in PA, especially UPA, potentially mitigated this loss, recovering up to 2.31 years (P < .01) overall. We explicitly acknowledge that our MET thresholds (LPA/MPA/UPA) are relative classifications derived from this cohort mortality risk profile. Although numerically distinct from guideline values, their sequence consistently reflects incremental intensity exposure; LPA corresponds to minimal activity, MPA captures light-to-moderate exertion, and UPA indicates sustained higher effort. Critically, the dose-dependent reduction in mortality risk and life expectancy recovery demonstrate biological coherence, independent of absolute MET values. This graded relationship – not specific thresholds – is the central finding with clinical relevance for multimorbidity management.

4.3. Sex-specific differences and novel contributions

A distinct finding deserving emphasis is the observed sex difference in the response to PA. Females exhibiting multimorbidity who engaged in PA appeared to recover significantly more life expectancy lost to multimorbidity than males (2.96 years vs 2.19 years). This suggests females with multimorbidity may derive greater survival benefits from PA, or require different intensity thresholds for benefit – a finding with potential clinical implications warranting further investigation in prospective studies. While the core protective effect of PA on mortality aligns with numerous studies,^[24,30–32] our detailed quantification of its impact specifically within a multimorbid population and the analysis of its modifying effect on life expectancy loss attributable to multimorbidity represent significant novel contributions. Furthermore, providing prevalence and outcome data across key subgroups (e.g., race/ethnicity, education, and BMI categories) offers critical granularity often lacking in broader population studies.

4.4. Methodological considerations and limitations

While utilizing a large, well-established national dataset is a major strength, several methodological aspects and limitations should be considered. First, disease classification relied on self-report and a limited set of lab tests. This approach, while standard in large surveys, could lead to misclassification bias (e.g., undiagnosed or subclinical conditions being missed), potentially influencing prevalence estimates, though likely less so the internal validity of the PA-mortality association if misclassification was non-differential by exposure status. Second, the restriction to 4 highly prevalent conditions (CVD, DM, HT, and RES) provides a focused analysis on major contributors to multimorbidity burden but inherently limits the comprehensiveness of our multimorbidity assessment. According to World Health Organization World Health Statistics 2025,^[33] CVD, cancer, DM, and RES constitute the 4 leading causes of premature mortality from NCDs, accounting for over half of NCD deaths among people under 70 globally. HT – a primary modifiable risk factor for CVD – amplifies this burden. While cancer is a major NCD, its heterogeneous subtypes and limited granularity in NHANES (e.g., inability to distinguish active/remitted cases or stage-specific effects) precluded meaningful inclusion. These conditions share underlying inflammatory and metabolic pathways directly amenable to PA interventions,^[34–36] strengthening the biological plausibility of our analysis. Future research including a broader array of chronic conditions (e.g., cancer, mental health disorders, arthritis) is essential for a more holistic picture. Third, our study population is drawn from US adult NHANES participants. While representative of the United States, this limits the direct generalizability of our specific prevalence rates and trends to other countries or healthcare systems with different demographics, risk factor profiles, and clinical practices. The underlying pathophysiological links between PA and reduced mortality risk, however, are widely accepted and likely applicable broadly. Fourth, although precise dose–response relationships (e.g., specific MET thresholds) for subgroups could not be established – and cross-study comparability is limited by cohort-specific thresholds – we mitigated these limitations by using RCS to define risk inflection points objectively, focusing conclusions on the exposure gradient (low to high intensity) over absolute values, and providing exact thresholds for reproducibility. Such detailed characterization should be a goal of future research. Finally, as an observational cohort study, despite adjustments, residual confounding remains a possibility.

4.5. Public health and clinical implications

Our findings significantly bolster the already compelling evidence underlying World Health Organization recommendations for regular PA (150–300 minutes MPA or 75–150 minutes UPA weekly) as a cornerstone strategy for chronic disease prevention and management.^[30,37] PA’s crucial role extends beyond initial prevention to mitigating the mortality risk and life expectancy loss associated with established multimorbidity. The persistent lack of global improvement in PA levels since 2001^[33] underscores the ongoing challenge. Designing and implementing effective, tailored, long-term PA promotion strategies requires addressing barriers specific to different populations, particularly high-risk groups identified in our study. Clinical practice must prioritize early detection and aggressive management of precursors like HT and DM to prevent or delay progression to multimorbidity. Crucially, for patients with existing multimorbidity, our data strongly support integrating structured, evidence-based PA counseling and support – including guidance on achievable intensity levels and self-management techniques tailored to individual capabilities and potentially differing efficacy by sex – as a core component of their clinical care plan.^[31,32] The substantial potential to recover lost life expectancy through PA represents a powerful motivator for both patients and providers.

5. Conclusions

This cohort study of US adults from 1999 to 2018 demonstrates that multimorbidity prevalence increased significantly over this period, imposing a substantial life expectancy penalty (approximately 5.6 years). CVD-based multimorbidity patterns were associated with the highest mortality risk. PA exerts a powerful protective effect, reducing all-cause mortality risk by approximately one-third compared to inactivity. Crucially, even among individuals with multimorbidity, moderate and especially vigorous PA significantly reduces mortality risk and recovers meaningful portions of lost life expectancy (up to 2.3 years overall). Our findings highlight a potentially greater life expectancy recovery benefit for females with multimorbidity engaging in PA compared to males. These results underscore the critical importance of PA as a modifiable factor for mitigating the substantial mortality burden associated with multimorbidity, supporting the integration of tailored, intensive PA promotion and support into public health initiatives and routine clinical management for all adults, particularly those at highest risk or already affected by multiple chronic conditions.

Acknowledgments

The authors thank all the members of the NHANES for their contributions and the participants who contributed their data.

Author contributions

Conceptualization: Dongdong Yu, Xiaoyu Cheng, Ting Cheng, Sihai Wang.

Data curation: Dongdong Yu, Yuxuan Shi, Tong Liu, Ting Cheng.

Investigation: Dongdong Yu.

Methodology: Dongdong Yu, Yongkang Wang, Xiaoyu Cheng, Ting Cheng, Sihai Wang.

Validation: Dongdong Yu, Yuxuan Shi, Huaying Fu, Tong Liu.

Visualization: Dongdong Yu, Yuxuan Shi, Huaying Fu, Tong Liu, Yongkang Wang.

Writing – original draft: Dongdong Yu, Yuxuan Shi, Xiaoyu Cheng, Ting Cheng, Sihai Wang.

Writing – review & editing: Dongdong Yu, Xiaoyu Cheng, Ting Cheng, Sihai Wang.

Formal analysis: Yuxuan Shi, Huaying Fu, Tong Liu, Ting Cheng.

Resources: Yongkang Wang.

Software: Yongkang Wang.

Funding acquisition: Xiaoyu Cheng, Sihai Wang.

Project administration: Xiaoyu Cheng, Ting Cheng, Sihai Wang.