We conducted a cross-sectional study using cluster random sampling to assess older adults in nursing homes. Based on economic classifications within Nantong City, four nursing homes were randomly selected as our survey sites. From March 3 to March 25, 2023, participants were recruited from the following nursing homes in Nantong City, Jiangsu Province, China: Nantong Yincheng Health Care Hospital, Nantong Sunshine Nursing Home, OlderYiyuan, and Nantong Sunshine New City Rehabilitation Nursing Home.

Our survey utilized the Wenjuanxing Software, a prominent online questionnaire platform. A trained researcher elucidated the study’s purpose to each participant before distributing our questionnaire link. Emphasizing anonymity, participants were encouraged to fill out the questionnaire autonomously, with the freedom to contact our researchers for any survey-related questions. Participants were also informed of their rights, including the option to withdraw from the survey at any point. Typically, it took most participants around 15 min to complete the questionnaire.

Eligible participants for the survey were those aged 60 years or older, residing in one of the four selected nursing homes, without major mental illnesses, and who were willing to provide written informed consent. Conversely, individuals who felt overwhelmed by the questionnaire and opted to discontinue midway or did not disclose their marital status were excluded. The cross-sectional survey formula was applied to calculate the sampling as follows:

Z_1−∂/2 is the statistical value for significance testing, where α = 0.05 and its value is 1.96. p is the prevalence rate of outcome variables, where q = 1−p and d is the allowable error, where d = 0.2p. Prior research indicates that the prevalence rate of outcome variables in older adults spans between 26.6% and 48.0% (27, 28). For our calculations in this study, we used the lower threshold of 26.6%, necessitating a sample size of at least 292 subjects when accounting for a 10% non-response rate.

Of the 602 older individuals surveyed, 506 questionnaires were fully completed, reflecting a response rate of 84.1%. Importantly, 118 older adults did not specify their marital status. Given our study’s focus on married and widowed individuals, these respondents were excluded from the final analysis. Consequently, our analysis encompassed 388 older adults: 239 married and 149 widowed. The average age of this cohort was 83.97 years, with a standard deviation (SD) of 7.19 years.

Ethical approval for this study was granted by the Ethics Committee of Nantong First People’s Hospital under the identification number 2023KT091.

As shown in Table 1, all participants belonged to the Han ethnicity. The sample included 242 females, accounting for 62.4% of the cohort. A majority, 298 older individuals (76.6%), were married. Furthermore, 117 participants (30.2%) reported a monthly disposable income in the range of 3,001–6,000 yuan. Notably, 127 (32.7%) of these older adults presented with varying levels of BMI abnormalities while residing in Chinese nursing homes.

In our PSM analysis, we designated married older individuals as the control group and compared them with their widowed counterparts in the treatment group. Figure 1 illustrates the propensity score distributions for both unmatched and matched samples. Notably, the total sample size diminished from 388 to 268, with 134 individuals each in the married and widowed categories, leading to the exclusion of 120 unmatched cases.

Before matching, the basic socio-demographic characteristics had an absolute standardized mean difference of 0.533. This difference decreased substantially to 0.016 after matching, indicating an excellent covariate balance in the post-matching samples (Figure 2). Additionally, chi-square and t-test results for these socio-demographic characteristics, categorized by marital status both pre- and post-matching, further validated the effectiveness of our PSM analysis. This efficacy was evidenced by the notable reduction in chi-square/T-value and p-value (Table 2).

Figure 3 delineates the interplay between marital status and the outcomes of mental health, PP, and fasting blood glucose among older residents of Chinese nursing homes, both prior to and subsequent to the matching process. In the pre-matching phase, after adjusting for socio-demographic covariates using multivariate linear regression models, widowed older individuals exhibited a notably elevated risk in PP (β = 8.00, p < 0.001), fasting blood glucose levels (β = 0.24, p = 0.039), and a reduction in perceived family support (β = −0.66, p = 0.003). Nevertheless, marital status did not exhibit a substantial link with either anxiety or depressive symptoms.

Post the PSM process, a commendable balance was achieved in all covariates between the married and widowed cohorts. The outcomes of the multivariate linear regression models in this phase diverged from their pre-matching counterparts, presenting augmented beta values. Notably, being widowed was significantly correlated with diminished family support (β = −0.81, p = 0.002), elevated depressive symptoms (β = 1.04, p = 0.043), increased PP (β = 8.90, p < 0.001), and heightened fasting blood glucose levels (β = 3.22, p = 0.027).

For a comprehensive understanding of how socio-demographic covariates impact the primary outcome variables, kindly refer to the Supplementary material provided.

We identified a notable link between widowhood and decreased family support among older individuals in nursing homes, though no significant relationship with anxiety symptoms emerged. This association emphasized the possible social and emotional challenges confronting widowed seniors, and this trend persisted in both pre-PSM and post-PSM analyses. Given the pivotal role of family in Chinese culture, a spouse’s loss can drastically alter the family support dynamics, subsequently affecting the individual’s well-being (31). Echoing prior studies (31–33), our data demonstrated a significant correlation between depressive symptoms and widowhood. It’s apparent that seniors living alone often lack regular interactions with family or friends, potentially escalating feelings of isolation and subsequent depression (34, 35).

While many studies highlight widowhood’s impact on anxiety (36–38), our findings aligned with alternative research, such as Sonja’s study in Zadar, suggesting no direct link between widowhood and anxiety (39). This divergence can arise from multiple factors, including the widowhood duration considered. The emotional repercussions and coping strategies may evolve with time after a spouse’s loss, affecting the onset of anxiety symptoms (40). To deepen our understanding, future investigations should consider the timeline of widowhood in relation to mental health trajectories.

Our findings indicated a strong link between widowhood and elevated PP in older adults nursing home residents (β = 8.90, p < 0.001). While prior studies have extensively studied the widowhood-hypertension connection with inconclusive results, the focus on PP has been limited. Liu’s research has highlighted widowhood as a significant hypertension risk in older Chinese individuals (41). Conversely, Randa’s study suggests that divorced or separated women have lower blood pressure compared to married women (40). Our study, pointing to a marked rise in PP in widowed individuals, echoed sentiments that emotional distress and lifestyle upheavals linked with widowhood can lead to cardiovascular complications (42). This reiterates the necessity for cardiovascular monitoring and interventions tailored for the widowed older adults.

On fasting blood glucose levels, we found a considerable association between widowhood and elevated glucose levels. This finding aligned with the idea of widowhood being linked to prediabetes (43). Potential dietary shifts, reduced physical activity, and heightened stress commonly experienced post-widowhood may drive metabolic disruptions (43, 44). Significantly, post-PSM linear regression revealed a heightened effect coefficient relative to pre-PSM analysis. By adeptly controlling confounding variables like age, sex, and BMI through PSM, we could precisely gauge the real impact of widowhood on fasting blood glucose levels while maintaining covariate equilibrium. This discovery stressed the imperative for interventions focused on diabetes management and prevention in widowed seniors.

Our study underscored the importance of PSM as an instrumental technique for managing covariate characteristics. By leveraging PSM, we enhanced the comparability between the married and widowed older residents in Chinese nursing homes, especially in relation to baseline socio-demographic attributes. Prior literature has emphasized the influence of such socio-demographic variables on the widowhood-health nexus in the older adults (45, 46). For instance, studies have indicated that widowed men exhibit higher depression levels than their female counterparts (45), while women seem to better acclimate to singlehood (46).

The ‘marriage resource model’ posits that marriage endows individuals with economic, social, and psychological benefits, which collectively uplift health and extend well-being (47, 48). This crucial facet, however, has often been overlooked in previous investigations. In our analysis, the initial absolute standardized mean difference for basic socio-demographic characteristics stood at 0.533. But with PSM application, this difference substantially narrowed to 0.016, signifying a notable covariate balance in our post-matching samples. The subsequent χ² and T values, which assessed socio-demographic attributes based on marital status pre- and post-matching, reinforced the efficacy of our PSM application in achieving this balance. This enhanced comparability allowed for a refined assessment in our linear regression models, where we observed distinct coefficients and beta values before and after matching. In essence, integrating PSM in observational research amplified statistical precision, bolstering the credibility and robustness of the derived conclusions.

There are several limitations to consider in this study. First and foremost, the cross-sectional nature of our data hindered the determination of causal links between the variables. To bolster our findings, subsequent studies should lean towards longitudinal designs, providing a more nuanced lens into the temporal relationships between variables. Second, while the use of PSM mitigated the influence of several socio-demographic variables, it’s crucial to recognize the potential existence of unmeasured confounders or residual confounding. Such factors might still wield influence over the relationships observed. Third, by relying on self-reported data, we inadvertently introduce the potential for biases, such as those stemming from social desirability or recall inaccuracies. Incorporating objective measures or gleaning data from multiple corroborative sources could elevate the credibility and robustness of findings in future research.