Long-term effectiveness of multidisciplinary weight management versus routine management on body mass index and waist circumference in older adults with overweight and obesity: a 24-month retrospective cohort study

Abstract

Background

Overweight and obesity pose an escalating health challenge for the older, yet routine management (RM) often underperform due to fragmented care. The multidisciplinary weight management (MWM) intervention addresses this challenge by integrating family doctor team and sport-health integration. We aim to evaluate the effectiveness of RM versus MWM intervention on body mass index (BMI) and waist circumference (WC) among overweight / obese adults (aged ≥ 65 years).

Methods

A 24-month retrospective cohort study utilized electronic health records from Xiamen Basic Public Health Cloud Platform, including the MWM group (1177 samples) and the Routine Management (RM) group (1188 samples). McNemar χ² tests were employed to assess within-group changes in BMI and WC from baseline to post-intervention, while χ² tests compared BMI and WC between groups. The Wilcoxon signed-rank test and the Mann-Whitney U test were utilised to compare clinical indicators between the two groups before and after intervention, as well as to compare the two groups. Multivariable linear regression models were fitted to report the effect of MWM versus RM as β coefficients with 95% confidence intervals (CIs). Mediation analysis were conducted on physical activity level, smoking status and alcohol consumption.

Results

The MWM group showed more favorable shifts in BMI (normal weight 13.25% vs. 0.00%) and WC (no centrol obesity 42.57% vs. 32.15%) compared with RM (P < 0.001). And MWM is associated with BMI and WC reduction, with a mean reduction in BMI of 0.88 kg/m² (95% CI: 0.77 to 0.99; P < 0.001) and in WC of 1.16 cm (95% CI: 0.78 to 1.55; P < 0.001). Physical Activity partially mediated the intervention effect on BMI (mediation proportion 6.78%; P = 0.032).

Conclusions

The MWM group produced clinically greater reductions in BMI and WC compared with RM, and its integrated care framework may offer an effective and scalable approach to obesity management in older adults. To promote this management, future research should validate in larger, diverse cohorts.

Graphical abstract

Study Flowchart and Analytical results for Comparing Multidisciplinary Weight Management (MWM) and Routine Management (RM) Groups

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-26509-y.

Background

The escalating global burden of obesity presents a particularly serious challenge to the health of aging populations. Older individuals with obesity face a heightened vulnerability to various metabolic conditions, including type 2 diabetes and cardiovascular disease. This risk is further compounded by the co-occurrence of sarcopenia, or age-related muscle loss, and excessive adipose tissue accumulation [1, 2]. Epidemiological projections indicate that by 2030, more than 20% of the world’s population aged 65 years and older will have obesity [3], increasing disability and death rates in this group. Older adults exhibit distinct body composition changes, wherein both excessively high and low BMI and WC are associated with elevated all-cause mortality. This U-shaped risk relationship underscores the necessity of balanced anthropometric management in geriatric populations management (RM) often implements weight management strategies insufficient for addressing the unique complexities of older adults, which include reliance on singular interventions neglecting synergistic physiological pathways, fragmented professional support contributing to poor adherence and weight regain, and insufficient integration of exercise medicine expertise for preserving musculoskeletal health [4–6]. This results in older individuals either having excessively high or low BMI and WC. This highlights the urgent need for better management frameworks for the geriatric population.

Contemporary research has therefore shifted toward multidisciplinary interventions that integrate dietary therapy, physical activity prescription, pharmacotherapy, and behavioral counseling, often supported by intelligent health-information systems and multidisciplinary team collaboration [7, 8]. Such approaches have been associated with favorable behavior changes (e.g., increased physical activity, reduced tobacco and alcohol consumption) and short-term weight reduction [9–11]. However, most trials report outcomes maintained for ≤ 6 months, longitudinal data beyond 24 months remain scarce, and evidence specifically addressing overweight and obese older adults is limited.

In 2021, Xiamen City launched an online and in-person Multidisciplinary weight management (MWM) intervention to prevent and treat overweight and obesity among the older. The intervention established a multidimensional health management service system (family doctor team and sports-health integration) by integrating the “Xiamen iHealth” platform (a hierarchical diagnosis and treatment system). It achieved seamless connectivity with primary care cloud platforms, Hospital Information System (HIS), and residents’ health records, enabling multidimensional interventions for chronic diseases (e.g. overweight and obesity). These actions highlight the critical role of comprehensive, coordinated interventions in ameliorating lifestyles and implementing refined, continuous management of chronic diseases. Detailed information about MWM can be found in Supplementary Material 1, Figure S1 and S2. Nevertheless, there is still a lack of knowledge about whether it can contribute to improved BMI and WC outcomes and serve as effective tools for illustrating the reduced obesity-related comorbidities within public health interventions.

Hence, this study evaluates the effects of the Xiamen MWM model on BMI and WC among community-dwelling older adults. We compared changes in BMI and WC between MWM and RM groups at follow-up to validate the model’s effectiveness. Subsequently, we assessed the influence of intervention-associated behavioral and lifestyle modifications on BMI and WC outcomes, comprehensively examining whether these changes mediate the model’s effect on weight management in this elderly cohort.

Methods

Data sources and study population

This retrospective cohort study utilized data from the Xiamen Basic Public Health Cloud Platform between 2021 and 2023. The platform integrates multiple databases, including Resident Electronic Health Records (EHRs), the Elderly Health Examination Information System, and the “Xiamen iHealth” family doctor contracted service application, enabling interoperability and data sharing. The study protocol received ethical approval from the Xiamen Center for Disease Control and Prevention Ethics Committee (Approval No.: XTK/LLSC [2023] 004). All participants signed written informed consent forms in accordance with the Basic Medical and Health Promotion Law of the People’s Republic of China (adopted at the 15th Session of the Standing Committee of the 13th National People’s Congress on December 28,2019, and implemented on June 1, 2020; details see Appendix A), with their records linked to personal medical insurance cards.

From 10 community settings in Xiamen, we enrolled 2,400 older adults who were overweight or had obesity (age ≥ 65 years; baseline body mass index [BMI] ≥ 24 kg/m²). Participants were ordered by health examination serial number and, according to enrollment in the family doctor contract service, classified into the multidisciplinary weight management group (MWM) and a routine management group (RM), with 1,200 individuals per group. Exclusion criteria were: mental disorders identified on screening scales (Ascertain Dementia 8-Item, AD8 and Community Screening Instrument for Dementia, CSI-D) and confirmed by higher-level medical institutions (n = 4); disorders of consciousness (n = 2); cognitive dysfunction (n = 2), and severe cardiac, hepatic, or renal insufficiency (n = 5); participation rate < 50% or loss to follow-up (n = 7); inability to continue participation due to physical or family circumstances (e.g., falls, relocation; n = 4); and other reasons leading to noncompletion of the study (n = 11).The final analytic cohort comprised 1,177 participants in the MWM group and 1,188 in the RM group (Fig. 1; Fig S3).

Fig. 1.

Study Design and Core Components of the Multidisciplinary Multidisciplinary Weight Management (MWM) Intervention Routine Management (RM) within the Xiamen Basic Public Health Services (BPHS) Framework. A cohort study (n = 2400) from Xiamen's public health platform compared Routine Management (RM) (n = 1200) and Multidisciplinary Weight Management (MWM) (n = 1200) excluding 35 samples per eligibility criteria

Definition of RM and MWM intervention

The RM intervention was delivered through quarterly centralized health education sessions focusing on weight management. These sessions utilized various formats including posters, brochures, lectures, consultations, and free clinic activities. All events were uniformly organized and announced by health educators. Participation was voluntary, with time slots provided but no mandatory attendance requirements. In contrast, the MWM intervention employed a dedicated family doctor team comprising general practitioners, specialists, and health managers who integrated exercise and health. Key features included personalized dietary counseling (e.g., creating individualized meal plans, maintaining daily food logs, lectures on reducing salt, oil, and sugar intake), as well as sessions covering weight, bone, and oral health. Interactive cooking demonstrations complemented these activities. Exercise guidance was tailored to individual prescriptions. Participants were encouraged to track daily activities, with the program incorporating exercise and health principles. Additional support involved sleep and psychological adjustments as part of a comprehensive lifestyle change strategy. Within this team framework, each member had specific roles: General practitioners handled early screening, initial diagnosis, basic treatment, health education, and long-term follow-up monitoring. The expert team (mostly from tertiary hospitals) led treatment optimization, provided professional consultation for complex cases, and promoted evidence-based individualized care through appropriate technologies. Health managers maintained electronic health records (EHRs), implemented lifestyle interventions, and offered medication and dietary guidance. Simultaneously, exercise guidance enhanced patients’ self-management capabilities. Routine monitoring is primarily conducted through the “Xiamen iHealth” mobile application, which enables users to track dietary and exercise habits. The app provides voice, video, text, and image communication features, along with online consultation services offering management advice. All interventions are systematically implemented throughout the 24-month study period. Face-to-face consultations and group activities are scheduled at least every three months, while online follow-ups or personalized consultations are arranged flexibly based on participants’ needs. For detailed procedures, see Fig S3 part III.

Assessment of body mass index (BMI), waist circumference (WC) and clinical indicators

In this study, the primary outcome measures were body mass index (BMI, kg/m²) and waist circumference (WC, cm) in older adults, while secondary outcome measures comprised clinical indicators. Anthropometric assessments include sphygmomanometers, stadiometers, digital weighing scales, and waist circumference tapes. All measurements were conducted by healthcare professionals who had received standardized training and followed a validated protocol. Standing height was measured using a TZG stadiometer (maximum range 2.0 m, minimum increment 0.1 cm), and body weight was recorded using a G&G TC-200 K/TC150KA electronic scale (minimum increment 0.01 kg, maximum capacity 150 kg). WC measurement employed a tape of identical brand and model (length 1.5 m, width 1 cm, minimum increment 0.1 cm), with the anatomical landmark defined at the midpoint between the lower border of the costal margin along the midaxillary line and the iliac crest, marked bilaterally for reproducibility. During measurement, the retractable tape was applied lightly against the skin, and readings were taken with the observer’s eyes aligned horizontally to the scale, recorded to the nearest 0.1 cm; duplicate measurements were taken, and values were recorded if the difference was less than 2 cm. Blood pressure was measured in a resting state using an Omron HBP1300/1320 digital sphygmomanometer (accuracy ± 1 mmHg). BMI categories followed the Working Group on Obesity in China (WGOC) criteria: underweight (< 18.5 kg/m²), normal weight (18.5–23.9 kg/m²), overweight (24.0–27.9 kg/m²), and obesity (≥ 28.0 kg/m²). WC was classified as normal (< 90 cm in male or < 85 cm in female) or central obesity (≥ 90 cm in male or ≥ 85 cm in female) [12, 13].

The clinical indicators measured in this study included fasting plasma glucose (FPG), whole blood lipid profile, complete blood count (CBC), urine analysis, and liver/renal function tests. Blood and urine samples were collected in the morning after participants had fasted for at least eight hours, during which they did not consume any caloric beverages or regular medications. All sample collection, processing, and storage were conducted under strictly controlled conditions following standard operating procedures. Laboratory analyses were performed at accredited facilities using automated analyzers of the same brand and model to ensure consistency. FBG was measured using validated methods including hexokinase or glucose oxidase assays. Lipid profiles included total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C), with TC and TG determined enzymatically and HDL-C/LDL-C measured directly. For CBC, EDTA-K₂-treated whole blood was analyzed using automated equipment, with manual microscopic verification when necessary. Urine analysis detected key indicators such as glucose and protein. Liver function tests followed international guidelines, measuring enzymes like alanine aminotransferase (ALT/AST), aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT), alkaline phosphatase (ALP), and bilirubin levels. Renal function was assessed through serum creatinine (Scr) and blood urea nitrogen (BUN). All results were double-checked by two reviewers and underwent consistency checks prior to entry into the health database. The annual physical examinations were scheduled at community health centers around January 15 each year, with a seven-day window period. Participants received advance notifications via WeChat or phone, as detailed in Fig S3 part IV.

Assessment of covariates and mediators

Based on previous studies, after excluding covariates with missing proportions > 20% [Proportion of missing data: albumin (1072/2365,45.33%), Bilirubin glucuronate (1160/2365, 49.05%) Blood Potassium (2138/2365, 90.40%), Blood Sodium (2170/2365, 91.75%), ect], the covariates included in this study comprised age, sex, education, marital status, ethnicity, living arrangement, pre-retirement occupation, income per month, hypertension, diabetes, baseline fasting blood glucose (FBG₂₀₂₁), baseline lipid profile, baseline complete blood count (CBC₂₀₂₁), baseline urinalysis, baseline pulse rate (PR₂₀₂₁) (beats/min), baseline Electrocardiogram (ECG₂₀₂₁), baseline liver function and baseline renal function. Details and standards are provided in Supplementary Material 2.

The mediating variables included in this study comprised physical activity level (calculated by using Adjusted Metabolic Equivalent of Task (MET_ad) values [Supplementary Material 3 to 5]), smoking status, and alcohol consumption [9–11]. Detailed information and standards are provided in Supplementary Material 2.

Statistical analysis

Descriptive statistics were used to report the baseline characteristics of the older study population. Continuous variables were presented as medians and interquartile range (IQR), while categorical variables were summarized as counts (percentages, %). The distribution of baseline characteristics was analyzed using the Mann-Whitney test for continuous variables and the χ² test or the Mann-Whitney test for categorical variables. A comparison of the changes in BMI, WC and some clinical indicators before and after the study was converted into categorical variables. These variables were then analyzed using the Wilcoxon signed-rank test or the McNemar χ² test. Subsequent to the study, a comparison was made of BMI, WC (BMI₂₀₂₃ and WC₂₀₂₃) and clinical indicators between the two groups. This comparison was made using the Mann-Whitney test or the χ² test.

Multivariable linear regression models (details in Supplementary Material 6) were used to estimate the intervention effects BMI or WC in older adults. The estimated values (β) and 95% confidence intervals (CIs) were calculated. In this analysis, three models were implemented: Model 1 was a crude model, including only the group variable (MWM and RM). Model 2 incorporated the group variable, age, sex, education level, marital status, ethnicity, living arrangement, pre-retirement occupation, income per month, hypertension, and diabetes. Model 3 extended Model 2 by including FBG₂₀₂₁, baseline lipid profile, CBC₂₀₂₁, baseline urinalysis, PR₂₀₂₁, ECG₂₀₂₁, baseline liver function, renal function, BMI₂₀₂₁, WC₂₀₂₁. Given that the Generalized Variance Inflation Factor (GVIF) values of the interaction terms between groups and statistically significant covariates exceeded 10, these interaction terms were not incorporated into the final model.

The “Mediation” packages were utilized to assess medication adherence, physical activity level, smoking status, and alcohol consumption. The average causal mediation effects (ACME), average direct effects (ADE), and proportion mediated (Prop.Med) were calculated based on 1000 resamplings using a nonparametric bootstrap approach based on percentile methods (Supplementary Material 7).

Sensitivity analyses using multivariable linear models were also conducted. We first used propensity score matching (PSM). A logistic regression model was built using the MWM and RM groups as the dependent variable and all baseline covariates as independent variables to generate a propensity score for each participant. Then, a 1:1 nearest-neighbor matching algorithm was applied with a calliper width set to 0.2 of the standard deviation of the logit of the propensity score to ensure matching quality. After matching, the balance of all covariates was assessed using absolute standardized mean differences (SMD), with an SMD of < 0.10 considered indicative of good balance [14]. Multivariable linear regression models were then repeated on the matched cohort to rigorously evaluate and minimize potential confounding bias from baseline covariates on the outcomes of interest (BMI and waist circumference at follow-up). Second, we stratified samples by BMI₂₀₂₃ subgroup (overweight or obese) and WC₂₀₂₃ subgroup (non-central obesity or central obesity) to evaluate the validity of the multivariable linear regression model. Third, subgroup analyses were conducted to evaluate the robustness of these findings. Subgroups were defined by age subgroup (65–79 years, ≥ 80 years), sex, marital status, ethnicity, living arrangement, pre-retirement occupation, education, income per month, hypertension, diabetes, Baseline lipid profile, CBC₂₀₂₁, baseline urinalysis, ECG₂₀₂₁, baseline liver function and renal function to evaluate the robustness of the multivariable linear regression model. The likelihood ratio test was employed to compare the interaction model with the non-interaction model, with the analysis of variance (ANOVA) method, and the Benjamini-Hochberg False Discovery Rate (FDR) P-value cutoff of 0.05 was used to indicate that whether the variable had an interactive effect on Group (MWM vs. RM). Finally, to quantify the required strength of association with both the group (MWM vs. RM) and the outcomes to fully explain away the observed effects, E-values were computed by the “EValue” packages after measuring BMI₂₀₂₃, WC₂₀₂₃, BMI₂₀₂₃ (after PSM), and WC₂₀₂₃ (after PSM) via multivariable linear regression.

R 4.5.1 was used for all statistical analyses. Multivariable linear regression was conducted using “lm” function. P < 0.05 was considered statistically significant.

Results

Study participants

This study included 2365 overweight or obesity older people (39.58% male; median [P₂₅, P₇₅] age, 69 [67, 73] years). Baseline demographic and anthropometric characteristics were compared between the Routine Management (RM, n = 1,177) and Multidisciplinary Weight Management (MWM, n = 1,188) groups for most variables, as detailed in Table 1. With the exception of statistically significant disparities in baseline characteristics, including ethnicity, education, and pre-retirement occupation (P < 0.05), no statistically significant variations were identified in the remaining baseline characteristics (P > 0.05).

Table 1.

Baseline demographic and anthropometric characteristics between routine management (RM) group and multidisciplinary weight management (MWM) group [n%/M (IQR)]

Variable		RM (n = 1188)	MWM (n = 1177)	|z|/χ2	P
Sex	Male	478 (40.24)	458 (38.91)	0.433	0.511
	Female	710 (59.76)	719 (61.09)
Age # (median)		69 (67 ~ 74)	69 (67 ~ 73)	1.475	0.140
Martial status	Married	1082 (91.08)	1090 (92.61)	1.849	0.174
	Others	106 (8.92)	87 (7.39)
Residency	Non-living alone	1079 (90.82)	1075 (91.33)	0.188	0.664
	Living alone	109 (9.18)	102 (8.67)
Ethnicity	Han	1099 (92.51)	1120 (95.16)	7.161	0.007
	Minority	89 (7.49)	57 (4.84)
Education #	Primary school or below/Junior high school	825 (69.44)	787 (66.86)	11.898	0.003
	Senior high school or technical secondary school	267 (22.47)	324 (27.53)
	College or Above	96 (8.08)	66 (5.61)
Pre-retirement occupation	Clerical	191 (16.08)	156 (13.25)	32.723	< 0.001
	Armed forces	75 (6.31)	61 (5.18)
	Agri-Forestry-Fishery	88 (7.41)	84 (7.14)
	Service & Sales	394 (33.16)	366 (31.10)
	Operators & Assemblers	180 (15.15)	210 (17.84)
	Managers	58 (4.88)	25 (2.12)
	Professionals	126 (10.61)	175 (14.87)
	Others	76 (6.40)	100 (8.50)
Income per month #	< ¥3000	199 (16.75)	152 (12.91)	0.901	0.368
	¥3000–4999	514 (43.27)	548 (46.56)
	¥5000–7999	284 (23.91)	307 (26.08)
	≥¥8000	191 (16.07)	170 (14.45)
Hypertension	No	399 (33.90)	376 (31.65)	1.358	0.244
	Yes	778 (66.10)	812 (68.35)
Diabetes	No	767 (65.17)	819 (68.94)	3.812	0.051
	Yes	410 (34.83)	369 (31.06)
BMI2021 subgroup	Overweight	992 (83.50)	990 (84.11)	0.162	0.687
	Obesity	196 (16.50)	187 (15.89)
WC2021 subgroup	No central obesity	455 (38.30)	467 (39.68)	0.472	0.492
	Central obesity	733 (61.70)	710 (60.32)
Baseline fast (FPG2021)#		5.99 (5.47 ~ 6.83)	5.96 (5.41 ~ 7.08)	0.121	0.904
Lipid profile (vs. normal)	Normal	378 (31.82)	370 (31.44)
	Hypertriglyceridemia (WHO: type I, IV)	102 (8.59)	99 (8.41)	1.807	0.771
	Hypercholesterolemia (WHO: type IIa)	121 (10.19)	135 (11.47)
	Mixed hyperlipidemia (WHO: types IIb, III, IV, V)	61 (5.13)	69 (5.86)
	Other types of Dyslipidemia	526 (44.27)	504 (42.82)
Baseline Pulse rate (PR2021)#		75 (69 ~ 79)	74 (68 ~ 80)	0.289	0.773
Baseline Complete Blood Count (CBC2021)	Normal	1070 (90.07)	1062 (90.23)	0.017	0.894
	Abnormal	118 (9.93)	115 (9.77)
Baseline urinalysis	Normal	673 (56.65)	699 (59.39)	6.051	0.195
	UP+/G+ (Glycosuria and Proteinuria positive)	30 (2.53)	19 (1.61)
	UP+/G− (Glycosuria negative and Proteinuria positive)	41 (3.45)	49 (4.16)
	UP−/G+ (Glycosuria positive and Proteinuria negative)	159 (13.38)	133 (11.30)
	Other Abnormal Urinalysis Indicators	285 (23.99)	277 (23.54)
Baseline Electrocardiogram status	Normal	488 (41.08)	471 (40.02)	0.276	0.599
	Abnormal	700 (58.92)	706 (59.98)
Baseline Liver function	Normal	874 (73.57)	874 (74.26)	0.145	0.703
	Abnormal	314 (26.43)	303 (25.74)
Baseline renal function	Normal	981 (82.58)	987 (83.86)	0.695	0.394
	Abnormal	207 (17.42)	190 (16.14)

① Hypertension is defined as three independent measurements on non-consecutive days showing systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg; ② Diabetes is defined as fasting blood glucose ≥ 7.0 mmol/L, 2-hour postprandial blood glucose ≥ 11.1 mmol/L, or glycated hemoglobin (HbA1c) ≥ 6.5%; ③ Lipid profiles are categorized into five groups based on normal ranges: high-density lipoprotein cholesterol (HDL-C) ≥ 1.0 mmol/L; low-density lipoprotein cholesterol (LDL-C) < 3.37 mmol/L), lipid profiles are categorized into five types: Normal (no abnormalities in all indicators); Hypertriglyceridemia (TG ≥ 1.7 mmol/L, WHO: Type I, IV); Hypercholesterolemia (TC ≥ 5.2 mmol/L, WHO: Type IIa); Combined Hyperlipidemia (TC ≥ 5.2 mmol/L & TG ≥ 1.7 mmol/L, WHO: Types IIb, III, IV, V); Other Lipid Abnormalities (abnormal HDL-C/LDL-C ratio). ④ Urinalysis indicators including proteinuria, glycosuria, ketonuria, and occult blood, with scoring ranges from “-” to “++++”. Any result of “±” or higher indicates abnormal urine parameters. Urinalysis results are categorized into 5 groups: Normal (all parameters within normal range), U^P+/G+ (positive for both glucose and protein), U^P+/G- (positive for protein, negative for glucose), U^P-/G+ (negative for protein, positive for glucose), Other types of abnormal urinalysis. ⑤ Complete Blood Count (WBC and PLT normal ranges: 4.0 × 10⁹/L to 10.0 × 10⁹/L and 100 × 10⁹/L to 300 × 10⁹/L, respectively), Liver Function (Serum Alanine Aminotransferase (ALT), aspartate aminotransferase (AST), and total bilirubin (TBIL), with normal ranges of 40 U/L, 40 U/L, and 17.1 µmol/L, respectively); renal function indicators (serum creatinine (Scr) and blood urea nitrogen (BUN) normal ranges: males 54–133 µmol/L, 44–97 µmol/L for females; normal BUN range: 2.8–7.2 mmol/L). Any value outside these ranges is considered abnormal

Variables labelled # are ranked or non-normal continuous variables and are tested using the Mann-Whitney test, while the rest of the variables are tested using χ² test. A two-tailed P < 0.05 was considered statistically significant, with significant results highlighted in bold

Changes and comparisons of BMI; WC and clinical indicators between the two groups

Comparative analysis revealed that the MWM intervention significantly outperformed the RM intervention in improving body mass index (BMI) and waist circumference (WC) (Table 2). The obesity prevalence in the MWM group decreased from 15.89% in 2021 to 0.68% in 2023, with 13.25% of previously overweight or obese individuals returning to normal BMI (P < 0.001). In contrast, the RM group saw obesity prevalence rise from 16.50% to 25.08%, with no participants achieving normal BMI (P < 0.001). Similarly, central obesity showed a decrease in prevalence from 60.32% to 57.43% in the MWM group (P = 0.030), while the RM group saw an increase from 61.70% to 67.85% (P < 0.001).

Table 2.

Changes and comparison of post-study body mass index (BMI₂₀₂₃) subgroups, waist circumference (WC₂₀₂₃) subgroups [n (%)]

Var	group	Categories of var	2021	2023	|z|/χ2	P	|z|post/χ2post	P post
BMI2023 subgroup	RM	Normal	0 (0.00)	0 (0.00)	7.478	< 0.001	20.942	< 0.001
		Overweight	992 (83.50)	890 (74.92)
		Obesity	196 (16.50)	298 (25.08)
	MWM	Normal	0 (0.00)	156 (13.25)	17.819	< 0.001
		Overweight	990 (84.11)	1013 (86.07)
		Obesity	187 (15.89)	8 (0.68)
WC2023 subgroup	RM	No central obesity	455 (38.30)	382 (32.15)	24.833	< 0.001	27.390	< 0.001
		Central obesity	733 (61.70)	806 (67.85)
	MWM	No central obesity	467 (39.68)	501 (42.57)	4.738	0.030
		Central obesity	710 (60.32)	676 (57.43)

The Wilcoxon signed-rank test was employed for BMI change, while McNemar's χ2 test was utilized for WC change. χ2 test was utilized for comparisons of WC between the two groups, whereas the Mann-Whitney U test was employed for comparisons of BMI. A two-tailed P <0.05 was considered statistically significant, with significant results highlighted in bold

After 24-month follow-up, the MWM group demonstrated significantly superior BMI and WC levels compared to the RM group: a higher proportion of participants restored to normal BMI (13.25% vs. 0.00%), lower obesity prevalence (0.68% vs. 25.08%), and lower central obesity prevalence (57.43% vs. 67.85%; all P < 0.001).

As for clinical indicators, baseline characteristics were comparable between the two groups (all P > 0.05). At the end of follow-up, the MWM group demonstrated significantly higher levels of high-density lipoprotein cholesterol (HDL-C) (P = 0.002), reduced triglycerides (P = 0.042), and improved hepatic function reflected by lower alanine aminotransferase (ALT) and aspartate aminotransferase (AST) values (both P < 0.001), as well as reduced total bilirubin (TBIL) (P = 0.045) and blood urea nitrogen (BUN) (P = 0.034), compared with the RM group. No significant differences were observed between groups in blood pressure, fasting blood glucose (FBG), heart rate, platelet count (PLT), or serum creatinine (Scr) (all P > 0.05). Within-group comparisons revealed significant reductions in total cholesterol (TC) and low-density lipoprotein cholesterol (LDL-C) in both groups (P < 0.002), while an increase in HDL-C was observed only in the MWM group (P = 0.024). Other results are shown in Table 3.

Table 3.

Comparison of clinical indicators between intervention and control groups [M (Q1–Q3)]

Clinical	Group	Baseline	Post-study	|z|a	P a	|z|b	P b	|z|c	P c
FBG (mmol/L)	RM	5.99 (5.47 ~ 6.83)	5.93 (5.44 ~ 6.96)	0.447	0.655	0.121	0.904	0.838	0.402
	MWM	5.96 (5.41 ~ 7.08)	5.98 (5.41 ~ 7.08)	0.502	0.616
TC (mmol/L)	RM	5.07 (4.30 ~ 5.78)	4.92 (4.19 ~ 5.70)	3.090	0.002	0.542	0.588	1.229	0.219
	MWM	5.05 (4.25 ~ 5.80)	4.89 (4.07 ~ 5.63)	3.727	< 0.001
TG (mmol/L)	RM	1.47 (1.08 ~ 2.00)	1.47 (1.12 ~ 2.02)	1.098	0.272	0.568	0.570	2.030	0.042
	MWM	1.48 (1.10 ~ 2.02)	1.41 (1.07 ~ 1.94)	1.631	0.103
LDL-C (mmol/L)	RM	2.97 (2.34 ~ 3.59)	2.84 (2.25 ~ 3.48)	3.615	< 0.001	0.290	0.772	0.874	0.382
	MWM	2.99 (2.31 ~ 3.60)	2.83 (2.14 ~ 3.47)	3.283	0.001
HDL-C (mmol/L)	RM	1.28 (1.10 ~ 1.51)	1.29 (1.09 ~ 1.53)	1.650	0.099	1.959	0.050	3.105	0.002
	MWM	1.31 (1.12 ~ 1.54)	1.32 (1.14 ~ 1.58)	2.260	0.024
SBP (mmHg)	RM	135 (128 ~ 145)	135 (128 ~ 144)	0.558	0.577	0.332	0.740	0.567	0.570
	MWM	135 (128 ~ 146)	135 (128 ~ 145)	0.829	0.407
DBP (mmHg)	RM	80 (74 ~ 85)	80 (74 ~ 85)	0.355	0.723	1.482	0.138	0.388	0.698
	MWM	79 (74 ~ 85)	80 (74 ~ 85)	0.711	0.477
PR (per min)	RM	75 (69 ~ 79)	75 (70 ~ 80)	3.907	< 0.001	0.289	0.773	0.557	0.577
	MWM	74 (68 ~ 80)	75 (69 ~ 81)	1.405	0.160
PLT (×109/L)	RM	226 (192 ~ 264)	229 (193 ~ 266)	1.630	0.103	1.171	0.242	1.723	0.085
	MWM	223 (191 ~ 259)	222 (189 ~ 262)	0.557	0.578
ALT (U/L)	RM	20.00 (16.00 ~ 26.20)	20.24 (15.50 ~ 28.00)	0.707	0.480	1.432	0.152	3.370	< 0.001
	MWM	20.00 (15.00 ~ 26.00)	19.00 (14.98 ~ 25.20)	1.150	0.250
AST (U/L)	RM	21.00 (18.00 ~ 25.00)	22.00 (18.00 ~ 26.00)	3.996	< 0.001	0.661	0.509	4.329	< 0.001
	MWM	21.00 (18.00 ~ 25.00)	21.00 (17.90 ~ 24.90)	0.213	0.832
TBIL (µmol/L)	RM	12.70 (9.99 ~ 16.00)	12.55 (10.10 ~ 16.02)	0.146	0.884	1.158	0.247	2.001	0.045
	MWM	12.80 (10.20 ~ 16.30)	13.00 (10.30 ~ 16.39)	0.179	0.858
Scr (µmol/L)	RM	70.00 (59.00 ~ 83.18)	71.00 (60.48 ~ 85.00)	4.125	< 0.001	0.666	0.505	0.956	0.339
	MWM	70.00 (59.00 ~ 82.00)	70.00 (60.00 ~ 84.20)	1.677	0.094
BUN (mmol/L)	RM	5.57 (4.66 ~ 6.51)	5.71 (4.79 ~ 6.76)	3.547	< 0.001	0.326	0.744	2.123	0.034
	MWM	5.56 (4.68 ~ 6.52)	5.54 (4.69 ~ 6.60)	0.800	0.424

^aComparison before and after intervention within groups (Wilcoxon signed-rank test), ^bComparison between groups at baseline (Mann–Whitney U test), ^cComparison between groups after intervention (Mann–Whitney U test). A two-tailed P <0.05 was considered statistically significant, with significant results highlighted in bold

Multivariable linear regression analysis of BMI or WC intervention effects

Multivariable linear regression analyses revealed that the MWM group achieved relevant reductions in BMI and WC compared with RM over a 24-month follow-up period (Table 4, S1-2). After adjustment for sociodemographic and baseline covariates, the MWM group exhibited a mean BMI reduction of 0.88 kg/m² (95% CI: −0.99 to − 0.77; P < 0.001) and a WC reduction of 1.16 cm (95% CI: −1.55 to − 0.78; P < 0.001) compared to the RM group. Key determinants included BMI₂₀₂₁ and WC₂₀₂₁ strongly predicting BMI₂₀₂₃ (β_BMI2021 = 0.65, 95%CI_BMI2021: 0.62 ~ 0.68; β_WC2021 = 0.02, 95%CI_WC2021: 0.01 ~ 0.03; all P < 0.001) and WC₂₀₂₃ (β_BMI2021 = 0.38, 95%CI_BMI2021: 0.25 ~ 0.50; β_WC2021 = 0.67, 95%CI_WC2021: 0.64 ~ 0.71; all P < 0.001). Further covariate analysis identified marital status (Others vs. Married: β = −0.22, 95% CI: −0.42 ~ − 0.02, P = 0.034 for BMI₂₀₂₃) and pre-retirement occupation (operators/assemblers vs. clerical: β = 0.86, 95% CI: 0.13 ~ 1.59, P = 0.021 for WC₂₀₂₃) as significant modifiers of BMI₂₀₂₃ or WC₂₀₂₃.

Table 4.

Multivariable linear regression of the effect of BMI₂₀₂₃ or WC₂₀₂₃ interventions

Outcome	Model	Categories of group			|t|	|t| (after PSM)	P	P (after PSM)
		RM	MWM (95% CI)	MWM (after PSM) (95% CI)
BMI2023	Model 1	Ref.	−1.23 (−1.37, −1.08)	−1.23 (−1.38, −1.08)	16.64	15.78	< 0.001	< 0.001
	Model 2	Ref.	−1.23 (−1.37, −1.08)	−1.23 (−1.38, −1.07)	16.40	15.66	< 0.001	< 0.001
	Model 3	Ref.	−0.88 (−0.99, −0.77)	−0.89 (−1.00, −0.77)	15.88	15.19	< 0.001	< 0.001
WC2023	Model 1	Ref.	−1.91 (−2.46, −1.36)	−1.85 (−2.43, −1.27)	6.80	6.28	< 0.001	< 0.001
	Model 2	Ref.	−1.88 (−2.42, −1.34)	−1.90 (−2.46, −1.34)	6.85	6.62	< 0.001	< 0.001
	Model 3	Ref.	−1.16 (−1.55, −0.78)	−1.16 (−1.56, −0.75)	5.91	5.61	< 0.001	< 0.001

The sample of study population with or without PSM is 2365 and 2174, respectively

Model 1 was adjusted for group

Model 2 was adjusted for age, sex, education level, marital status, ethnicity, living arrangement, pre-retirement occupation, monthly income, hypertension, diabetes

Model 3 was adjusted for age, sex, education level, marital status, ethnicity, living arrangement, pre-retirement occupation, monthly income, hypertension, diabetes, baseline fasting blood glucose, lipid profile, complete blood count, urinalysis, pulse rate, ECG, liver function and renal function, BMI₂₀₂₁, WC₂₀₂₁

BMI₂₀₂₃,post-study body mass index; WC₂₀₂₃,post-study waist circumference. |t|, t values, Ref., Reference

P value was calculated across group using multivariable linear regression models. A two-tailed P < 0.05 was considered statistically significant, with significant results highlighted in bold

Meditation analysis

Mediation analysis (Fig. 2) revealed that the MWM group yielded significantly greater reductions in BMI and WC compared to routine management (RM) over 2 years, with a notable Average Direct Effect (ADE: −0.9573, − 1.3388, all P < 0.001). Importantly, physical activity significantly mediated BMI₂₀₂₃, with an average causal mediation effect (ACME: 0.0607; 95% CI: 0.0064, 0.1184) and a proportion mediated of −6.78% (P = 0.026). In contrast, smoking status and alcohol consumption did not demonstrate significant mediating effects.

Fig. 2.

Mediating analysis results of medication adherence, physical activity level, smoking status and alcohol consumption in the association between group and BMI and WC intervention. Mutivariable linear regression analysis were adjusted for age, sex, education, marital status, ethnicity, living arrangement, pre-retirement occupation, income per month, hypertension, diabetes, baseline fasting blood glucose, lipid profile, complete blood count, urinalysis, pulse rate, ECG, liver function and renal function, BMI₂₀₂₁, WC₂₀₂₁. Solid arrows denote average direct effects (ADE) (e.g., the direct influence of Group on the outcome, independent of mediators), while dashed arrows represent average causal mediation effects (ACME) (e.g., the indirect pathway through which Group affects the outcome via a mediator). All reported ACME, ADE, confidence intervals (CIs) for these effects were constructed using the nonparametric bootstrap method with the percentile approach, based on 1,000 resamples

Other sensitivity analyses

After implementing propensity score matching (PSM) to minimize potential confounding from baseline characteristics, a total of 1057 pairs of participants from the RM and MWM groups were successfully matched (Table S3, Fig.S2). The MWM group exhibited a mean reduction in BMI of 0.89 kg/m² (95% CI: 0.77 to 1.00; P < 0.001) and a reduction in WC of 1.16 cm (95% CI: 0.75 to 1.56; P < 0.001), compared to the RM group. When compared with the results obtained without PSM, the errors of the above outcomes were all within 5%, though some baseline data are statistically different.

The results (Tables S4 and S5) demonstrate the MWM’s robust intervention effect compared to RM on BMI₂₀₂₃ and WC₂₀₂₃. Stratified multivariable linear regression analyses consistently demonstrated significant reductions in BMI across nearly all demographic and clinical subgroups, with effect sizes ranging from 0.66 to 2.23 kg/m². Similarly, significant reductions in waist circumference (WC) were observed across most subgroups, ranging from 0.57 to 5.84 cm.

Notably, interaction tests using likelihood ratio tests (Table S6) indicated that the intervention effect on BMI reduction was significantly modified by baseline BMI/WC subgroups (FDR P < 0.001) and baseline ECG (FDR P = 0.027). Similarly, the effect on WC was modified by baseline BMI/WC (P < 0.001), liver function (FDR P = 0.030). Subgroup analyses elucidated these interactions: individuals with combined obesity and central obesity derived substantially greater benefit from the MWM intervention [β = −2.23 (95%CI: −2.66 ~ − 1.79) for BMI; β = −4.22 (95%CI: −5.73 ~ − 2.70) for WC] compared to other subgroups.

After further adjusting for FBG₂₀₂₁, baseline lipid profile, CBC₂₀₂₁, baseline urinalysis, PR₂₀₂₁, ECG₂₀₂₁, baseline liver function and renal function, the results of the sensitivity analysis remained consistent. However, after additional adjustment for BMI₂₀₂₁ and WC₂₀₂₁, we still observed the effective intervention of BMI and WC in the MWM group, though the reduction was reduced.

E-values (Table S7) indicated that unmeasured confounding was unlikely to fully explain the intervention effects. The E-value for BMI₂₀₂₃ was 2.99 (lower 95%CI: 2.73), 3.01 (lower 95%CI: 2.74) for BMI₂₀₂₃ (after PSM), 1.61 (lower 95%CI: 1.45) for WC₂₀₂₃ and 1.60 (lower 95%CI: 1.43) for WC₂₀₂₃ (after PSM), suggesting greater robustness for the BMI₂₀₂₃. These results indicate that the estimated beneficial effect of the MWM group on BMI is relatively robust to unmeasured confounding, given the considerable strength of confounding required to explain the effect. The effect on WC, while still requiring non-trivial confounding to be fully explained, appears somewhat more susceptible to potential unmeasured variables.

Discussion

This 24-month multidisciplinary weight management (MWM) intervention, through the integration of family doctors, sport-health integration, and the Xiamen iHealth APP, significantly improved body mass index (BMI) and waist circumference (WC) in elderly participants. These findings not only confirm existing long-term evidence on obesity management in older adults but also expand upon it. Notably, the magnitude of BMI change in our cohort exceeded the reduction reported in a comparable 24-month, multi-component intervention trial (BMI decrease of 0.50 kg/m² and WC decrease of 2.10 cm) [15], further demonstrating its efficacy. The observed superiority may be attributed to MWM’s unique three-tier integrated care structure with family doctor team, which comprehensively addresses the multifactorial pathophysiology of obesity through clinical treatment, nutritional counseling, and personalized exercise prescriptions under supervision [16–18]. It is worth noting that our framework integrates sport and health through customized physical activities [19], partially making up for the deficiency of traditional nursing intervention measures [20]. Furthermore, the 24-month sustained effect demonstrates that MWM intervention supports long-term adherence, overcoming the common limitation of short-term interventions [21].

Interestingly, while numerous studies have reported that obesity and central obesity impair weight management outcomes, stratified analyses reveal that individuals with higher baseline BMI and WC achieved more significant benefits during interventions. This suggests that prioritizing high-BMI and high-WC populations in lifestyle intervention programs may optimize public health outcomes [22]. However, we also notice that the WC threshold used in this study was lower than internationally accepted standards, reflecting variations in body fat distribution and associated risk factors among East Asian populations. This adjusted threshold enhances the sensitivity of abdominal obesity detection, which may significantly influence cross-population comparisons and global policy frameworks [23].

Mediation analysis revealed that among behavioral factors assessed (physical activity, smoking, and alcohol consumption), only increased physical activity level demonstrated statistical significance as a mediator for BMI reduction. Both BMI and WC exhibited substantial average direct effects (ADE), indicating that most observed benefits were independent of the measured behavioral mediators. This suggests potential mechanisms may involve: (1) enhanced metabolic regulation and reduced systemic inflammation through integrated physiological pathways; or (2) synergistic effects of multi-component interventions, with their impact extending beyond individual lifestyle variables [24]. These findings underscore the complexity of multidisciplinary interventions, where the integrated care structure itself may generate holistic improvements beyond the sum of its individual components. Nevertheless, given that this study is a retrospective study, other lifestyle indicators such as dietary patterns or sleep quality cannot be included. This suggests that future prospective studies should include a wider range of behavioral and biological markers to better elucidate these pathways.

Furthermore, interventions targeting overweight and obesity indicators should also consider demographic and clinical factors [25]. Subgroup analysis revealed significant heterogeneity in intervention efficacy across different demographic and clinical characteristics. The study demonstrated that Marital status (others) and those with baseline diabetes showed greater reductions in BMI, while participants with pre-retirement obligation (Operators & Assemblers) or hypertension exhibited relatively smaller improvements in WC. This may be related to metabolic changes associated with occupation or physiological disorders of visceral adipose tissue reduction [26–29]. Despite this heterogeneity, propensity score matching (PSM) demonstrated that the intervention effect remained robust after adjusting for baseline variables. E-value analysis indicated that unmeasured confounding was unlikely to fully account for the observed benefits.

This study innovatively conducted a 24-month follow-up, comprehensively adjusted for demographic and clinical covariates, and employed causal mediation analysis combined with E-value sensitivity testing to assess residual confounding factors. These methods enhanced the internal validity of our conclusions regarding the effects of MWM intervention. However, this study has several limitations that require consideration: First, as an observational retrospective study, the causal relationship cannot be definitively established [30], and the the grouping method based on family doctor contracted services may introduce selection bias. Second, self-reporting behavior measurement may affect the generalizability of the results [31]. Third, The lack of physiological or biochemical assessments (such as visceral adipose tissue area, metabolic syndrome components, and inflammatory cytokine profiles) limits conclusions about long-term sustainability [32]. Finally, more sophisticated machine learning models could be used in future studies to replace multivariable linear regression, to improve predictive and interpretive power [33]. Future studies should therefore incorporate these components and cover more diverse populations to validate and optimize.

In conclusion, Xiamen City has implemented a MWM intervention involving family doctor team and sport-health integration. Leveraging internet technology, this intervention combines online and in-person methods to enhance management efficiency and patient adherence, which effectively addresses overweight and obesity issues among the older people. The intervention holds potential for further exploration and expansion to more communities and other regions, facilitating deeper integration of weight management with chronic disease prevention and control.

Authors’ contributions

Y.C. and L.H. conceived and designed the study, and drafted the manuscript. Y.L. and Y.H. performed the feasibility assessment, and were involved in drafting and revising the manuscript. H.Z. conducted the literature search, data collection, and data curation. L.H. carried out the analysis and interpretation of the results. Z.G. was responsible for quality control, critical revision, supervision, and acquisition of funding for the cohort study. All authors critically reviewed and approved the final manuscript.

Funding

This work was supported in part by the Project on the Establishment and Utilization of an Elderly Health Examination Cohort and a Major Chronic Disease Registry (Grant No. JKCLPJ202501002).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval for this research was granted by the Institutional Review Board of the Xiamen Center for Disease Control and Prevention, China [Approval No. XTK/LLSC(2023)004]. The study was conducted in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments, as well as relevant national and institutional guidelines and regulations. Prior to participation, all individuals were provided with a comprehensive explanation of the study objectives and procedures. Written informed consent was obtained from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.