Patterns of CPBMI-OC and associated factors among platform workers in new forms of employment in China: a cross-sectional study

Abstract

Introduction

Despite the basic medical insurance system achieving 95% coverage in China, platform workers in new forms of employment (PWNFEs) face significant challenges in maintaining continuous participation in basic medical insurance (CPBMI). This study aims to identify distinct patterns of CPBMI by occupational characteristics (CPBMI-OC) and their associated factors.

Methods

We conducted a cross-sectional survey with 641 PWNFEs in China using a structured questionnaire and employed latent class analysis (LCA) to identify patterns of CPBMI-OC. We utilized multinomial logistic regression to examine the associations between patterns of CPBMI-OC and variables, including demographics, socioeconomic status, medical service utilization, and health and insurance statuses.

Results

Among the 641 PWNFEs surveyed (74.4% aged 20–39 years, 79.4% male), the basic medical insurance (BMI) coverage rate was 38.7%, with 85.6% of participants reporting interruptions. LCA identified three distinct patterns: (1) fully interrupted, high-income, family migration class (13.4%); (2) high interruption, mid-income, stable residence class (30.9%); and (3) high continuity, low-income, non-contracted individual mobility class (50.7%). Urban and Rural Resident Basic Medical Insurance and urban hukou were positively associated with higher CPBMI probability. In contrast, higher education, better self-rated health, female gender, supplementary insurance coverage, and platform-insurance enrollment linkage were associated with a higher interrupt probability of CPBMI.

Conclusion

This study highlights the low CPBMI rate. The diverse patterns of CPBMI-OC among PWNFEs underscore the systemic challenges associated with flexible BMI enrollment options, which hinder continuous insurance participation. Our findings emphasize the need for targeted policy interventions to address structural inequities, improve the inclusiveness of BMI schemes, and better accommodate the diverse needs of PWNFEs.

Introduction

Worldwide, platform workers in new forms of employment (PWNFEs) show vulnerability in their ability to maintain continuous participation in basic medical insurance (CPBMI). This phenomenon fundamentally stems from their unique occupational characteristics (OC), which include shift work patterns, various job types, ambiguous contractual relationships, dynamic work locations and variable durations [1]. These OC expose PWNFEs to elevated health risks and increased medical costs [2] and contribute to volatility and uncertainty in their insurance participation [3,4,5]. Consequently, OC in health insurance coverage emerges as a critical determinant of health disparities among labor populations [6]. Given these complexities, it becomes imperative to investigate how occupational characteristics shape continuous participation in basic medical insurance (CPBMI-OC) patterns among PWNFEs.

In developed countries where informal economies bloom, social health insurance system has been adapted to provide relatively stable coverage for ensuring access to healthcare and protecting PWNFEs [7], while self-employed and low-income contracted PWNFEs still face challenges in accessing employer-sponsored health insurance [8, 9]. On the other hand, in many developing countries, numerous studies reveal that PWNFEs have low health insurance coverage rates [10, 11]. In China, the basic medical insurance (BMI) system comprises two main schemes: the Urban Employee Basic Medical Insurance (UEBMI) for formal employees and the Urban and Rural Resident Basic Medical Insurance (URRBMI) for informal workers [12]. The BMI system had almost reached 95% coverage by the end of 2023, with 1.33 billion participants (370.95 million covered by UEBMI and 962.94 million by URRBMI) [13]. The All-China Federation of Trade Unions reported 84 million PWNFEs, accounting for 21% of the labor force, and the majority engaged in ride-hailing drivers, couriers, food delivery, and webcasting. Current insurance policies remain insufficient for the needs of this group [14, 15]. Their flexible insurance options (choosing UEBMI or URRBMI), high mobility, large group size, complex occupational structures, and ambiguous labor relations [16], pose unique challenges in accessing the BMI system.

Conventional methods for analyzing determinants, such as ordered Probit and multinomial Logistic regression, have been widely used to explore factors influencing insurance participation behavior [17, 18]. These studies have highlighted institutional factors, knowledge of the insurance system, health status, socioeconomic status (SES), and demographic characteristics as key determinants of insurance participation among PWNFEs [14, 19,20,21,22]. Furthermore, their insurance enrollment is influenced by types of jobs, mobility frequency, labor contract status, and income disparities [23, 24]. However, most studies examined participation in insurance as a binary outcome (enrolled or not) or focused on comparisons between different types of insurance. Conventional methods for determinants analysis are limited in capturing the complexity of CPBMI-OC, an alternative approach is needed.

In the field of occupational health, Latent Class Analysis (LCA) can effectively identify distinct work characteristic clusters within and across OC [25]. For instance, Davalos [26] used LCA to classify workers into “voluntary” and “non-voluntary” informality, while Wright [27] applied LCA to explore patterns of informal domestic workers’ working environments based on economic and social characteristics. Thus, we aim to use LCA to identify distinct patterns of CPBMI-OC among Chinese PWNFEs and then examine associated factors with demographics, SES, health status, insurance status, and medical services utilization. Identifying factors affecting differences in CPBMI-OC patterns might enhance health protection and insurance equality among PWNEFs in China.

Methods

Study population and data collection

The research subjects included PWNFEs recruited online and offline. Online participants relied on the Wenjuanxing platform, and offline participants came from the Meituan Company and the Tianjin Housekeeping Service Company. The questionnaire was designed for this study to capture patterns of CPBMI-OC and associated factors among PWNFEs in China (Supplementary 2). The questionnaire included demographic characteristics, SES, health status, insurance status, medical services utilization, and occupational factors. A total of 756 questionnaires were collected, and the inclusion criteria were (1) age ≥ 18 years, (2) have worked on the platform for a minimum of one month, (3) provide informed consent, and (4) questionnaire completion > 120 s. Questionnaires with missing data were also excluded from the analysis. Ultimately, 641 participants were included for descriptive analysis and identification of continuous insurance participation behavior patterns related to occupational characteristics.

Patterns of CPBMI-OC

We assessed CPBMI by determining whether participants have maintained uninterrupted BMI enrollment since initial registration (continuous = 1, interrupted = 0). We also recorded the type of BMI currently enrolled and categorized participants without insurance as interrupted. To better understand the structural barriers and enablers of CPBMI, we examined patterns of CPBMI-OC, including employment contract status, careers, labor mobility, monthly income, and employment duration. These characteristics, combined with CPBMI, contribute to distinct CPBMI-OC profiles.

Employment contract status was assessed by asking respondents, “Have you signed a formal labor contract with the platform? " We also surveyed the specific careers in which respondents were engaged. According to previous studies [28, 29], we classified these careers into labor-oriented occupations (online task crowdwork, including food delivery and courier services, ride-hailing or chauffeur services, intra-city freight transport, domestic services, security work, and express logistics) and creative-oriented occupations (‘playbour’ crowdwork, including webcasting and online novel writing, and profession-based freelance crowdwork, such as platform-based self-employment). Labor mobility was determined through two questions: (1) whether the insurance registration area matched the workplace location (yes vs. no) and (2) if the registration area did not match the workplace, whether the respondent migrated alone or with family members. Participants who did not migrate for employment were coded as 1, those who relocated alone were classified as individual (value = 2), and those who migrated with family members (e.g., spouse, children, or parents) for employment were classified as ‘With family’ (value = 3). Based on monthly income levels, we categorized income into four groups: 1 (< 4,000 CNY), 2 (4,000–8,000 CNY), 3 (8,000–12,000 CNY), and 4 (> 12,000 CNY). Employment duration reflects employment stability and the length of time workers have integrated into their local community. We classified respondents into four groups: 1 (Stable residence), 2 (< 1 year), 3 (1–5 years), and 4 (5–10 years). Using LCA, we analyzed the above variables to identify their association with CPBMI patterns.

Potentially associated factors

Guided by the literature [23, 30,31,32,33,34], we included a range of independent variables to examine the potentially associated factors. For demographics, we included age (20–39 years or 40–59 years) and gender (male vs. female). SES included hukou status (rural vs. urban) and education (secondary and lower vs. tertiary and above). Insurance status encompassed the type of BMI categorized as uninsured, UEBMI, and URRBMI. We also assessed the availability of supplementary insurance (yes vs. no), based on whether the respondent had enrolled in commercial health insurance or Huiminbao (a low-cost, inclusive supplementary health insurance plan). Health status was measured by self-rated health (SRH) and chronic conditions. SRH is a reliable indicator of general health, as respondents tend to provide consistent ratings when their health remains unchanged [35]. We categorized scores of 0–6 as poor health, 7–8 as moderate health, and 9–10 as good health for analytical simplicity. We evaluated the chronic condition by asking, “Have you received a diagnosis for any of the following chronic diseases from a doctor?” The list included hypertension, coronary heart disease, diabetes, stroke, chronic obstructive pulmonary disease (COPD), and other chronic illnesses. Medical service utilization included past-year hospitalization (yes vs. no) and outpatient visits in the past two weeks (yes vs. no).

Statistical analysis

We conducted LCA on 641 participants using five occupational characteristics to identify distinct patterns of CPBMI-OC. LCA is a method for uncovering hidden subgroups within a dataset by analyzing response patterns across indicator variables [36]. It puts individuals into certain classes based on posterior membership probabilities, which gives us information about how people behave [37].

We examined classes one to five and selected the best-fitting solution based on an evaluation of various model-fitting statistics. The model fit was assessed by BIC (Bayesian Information Criterion), AIC (Akaike Information Criterion), aBIC (adjusted Bayesian Information Criterion), Log (L), LMR (Vuong-Lo-Mendell-Rubin likelihood ratio test), and BLRT (Bootstrapped likelihood ratio test) indicators. For the AIC, BIC and aBIC, a lower score is indicative of model-fit statistic and other tests indicates if one model is statistically better than another [38]. Additionally, entropy was calculated to gauge classification precision, with higher entropy indicating greater accuracy in subgroup assignment. We adopted BIC as the primary indicator of the final solution, as it is widely recognized as the most reliable indicator in LCA [39]. After choosing the best LCA model, we found out what pattern each respondent classed to and used that as the dependent variable in a multinomial logistic regression analysis to look at the factors that affected the patterns we identified. This model included demographic characteristics, socioeconomic status (SES), health status, insurance status, and medical service utilization as independent variables. Variables assignments are shown in Supplementary (eTable 1). Variance inflation factor (VIF) was used to assess the severity of multicollinearity, and VIF greater than 10 is considered indicative of severe multicollinearity [40]. All analyses used the ‘poLCA’ [41] and ‘nnet’ [42] packages in R 4.4.1. Statistical significance was P < 0.05 and α = 0.05.

Results

Sample basic characteristics

Table 1 shows the characteristics of the 641 participants used for the LCA. Of all participants, 477 (74.4%) were aged 20–39. There was a higher proportion of males (79.4% vs. 20.6%). In terms of SES, 384 (59.9%) had urban hukou, 298 (46.5%) had a monthly income of 4,000–8,000 CNY, 154 (24.0%) earned 8,000–12,000 CNY, and 131 (20.4%) earned less than 4,000 CNY per month. The majority (559, 87.2%) engaged in labor-based occupations, and others worked in creative-based occupations. Employment contract status revealed that 180 (28.1%) had signed formal contracts, while 461 (71.9%) lacked formal contracts. The BMI coverage rate was low at 38.7%, with 92 (14.4%) insured by UEBMI and 156 (24.3%) covered by URRBMI; 77 (12.0%) had supplementary health insurance. Only 92 (14.4%) continued to participate in BMI, and 549 (85.6%) reported interrupting CPBMI.

Table 1 Demographic characteristics of latent classes of the CPBMI-OC pattern

LCA of CPBMI-OC and model fit results

To identify the most appropriate CPBMI-OC patterns, we conducted LCA to compare models with one to five classes (Supplementary eTable 2). The model-fit statistics varied across different classes. The one-class model showed a log-likelihood test (P < 0.001) suggesting correlations among occupational characteristics, and additional latent classes were needed to improve model-fit. The two-class model showed better fit than the one-class model (entropy = 1.000), however, its AIC, BIC and aBIC were higher than the three-class model. The three-class model had the lowest BIC (5585.382), other fit statistics were also favorable (entropy = 0.898, AIC = 5429.176, aBIC = 5474.260). LMR and BLRT tests were statistically significant (P < 0.001), supporting the three-class model better than the two-class model. Although the four-class model AIC and aBIC were slightly lower than three-class model, one of its classes contained less than 10% of the sample, which suggests potential small class sizes and complicates interpretation [36]. The five-class was not supported, with LMR test result (P = 0.0984). After the trade-off between model-fit statistical, classification quality, and interpretability, the three-class model was selected as the optimal solution for identifying CPBMI-OC patterns.

In Fig. 1, we named the three CPBMI-OC patterns: fully interrupted, high-income, family migration class (Class 1, 13.4%); high interruption, mid-income, stable residence class (Class 2, 30.9%); and high continuity, low-income, and non-contracted individual mobility class (Class 3, 55.7%).

Fig. 1

Three-class model of CPBMI-OC pattern

Class 1 was characterized by a 100% CPBMI interruption rate, despite higher incomes (50.8%, > 8000 CNY per month); 64.0% migrated with their families for employment, and 69.9% reported signing formal employment contracts. Class 2 exhibited a high rate of interrupted CPBMI at 90.9%. They primarily belonged to the stable and mid-income group with a more balanced income distribution group (39.9%, 4000–8000 CNY per month). 53.5% reported signing formal employment contracts, and none reported employment-related migration. In contrast, Class 3 had the highest CPBMI rate (21.8%), despite having the lowest income (78.3%, < 8000 CNY per month). Individuals in Class 3 had labor-based jobs, lacked formal contracts (99.4%), and displayed high individual mobility (55.3%).

Associated factors of CPBMI-OC patterns

Table 1 shows that univariate analysis indicated differences in demographics, SES, health status, and medical service utilization among CPBMI-OC classes. Compared to participants in Class 1, those in Class 2 included fewer females (36.4%), were more likely to register in rural hukou (71.2%), and had a higher proportion of individuals with supplementary health insurance (25.8%). In Class 3, participants were older (31.1%, aged 40–59 years), predominantly male (94.4%), less educated (87.4% with secondary education or lower), and exhibited a higher proportion of urban hukou status (82.9%). This group also had the highest percentage of participants enrolled in URRBMI (35.6%) and reported their health as good (52.4%), whereas Class 1 had the highest proportion of poor health ratings (27.9%). Additionally, Class 3 participants utilized fewer medical services than Classes 1 and 2. There were no significant differences in chronic conditions across the classes (P = 0.118).

The univariate analysis observes that several demographic characteristics, SES, health status, insurance status, and medical service utilization factors are associated with CPBMI-OC patterns among PWNFEs (Supplementary eTable 3). Table 2 further explores these associations and provides the relative risk ratio (RRR) and 95% confidence interval (CI) for each variable associated with CPBMI-OC patterns in multinomial logistic regression. Severe multicollinearity was not detected as all VIFs were less than 10. Gender showed significant associations. Females were significantly less likely to be in Class 3 than males (RRR = 0.12, 95% CI: 0.06–0.25, P < 0.001). SES, including higher education and having an urban hukou, were negatively associated with Class 2. Participants with tertiary and over education (RRR = 0.48, 95% CI: 0.25–0.92, P = 0.027) and those having an urban hukou (RRR = 0.52, 95% CI: 0.28–0.96, P = 0.036) were significantly less likely to be in Class 2. Higher education also reduced the likelihood of being classified in Class 3 (RRR = 0.12, 95% CI: 0.06–0.24, P < 0.001), but urban hukou status was positively associated with this class (RRR = 2.95, 95% CI: 1.51–5.75, P = 0.002). Insurance status was also found to be an important influencing factor. Participants insured in URRBMI or those without supplementary health insurance had a significantly higher probability of being in Class 3 (compared with Class 1), with RRRs of 3.88 (95% CI: 1.43–10.50, P = 0.008) and 4.23 (95% CI: 1.53–11.7, P = 0.005), respectively. Furthermore, participants who linked their platform work registration to insurance coverage had a significantly lower likelihood of falling into Class 2 than Class 1, with an RRR of 0.12 (95% CI: 0.01–0.92, P = 0.041).

Table 2 Multinomial logistic analysis of factors associated with the CPBMI-OC patterns

SRH emerged as an important factor influencing CPBMI-OC patterns. Moderate SRH showed a significant negative association with Class 3 (RRR = 0.41, 95% CI: 0.18–0.91, P = 0.028). However, chronic conditions had no significant influences on CPBMI-OC patterns (P = 0.062 for Class 2; P = 0.119 for Class 3). Additionally, we observed no significant associations in age, medical service utilization, and the patterns under study.

Discussion

Previous studies have analyzed OC without considering the combined effects and interactions of multiple occupational factors on CPBMI. Our study found insurance continuity varies across different OC. Specifically, occupational contracts do not enhance insurance continuity for PWNFEs, LCA reveals that PWNEFs without a formal contract tend to exhibit higher continuity. This suggests that platform companies have not sufficiently addressed the insurance needs of these PWNEFs, who may be particularly eager to receive support from platform employers. There are variations between family and individual mobility on CPBMI. PWNEFs who have experienced family mobility are more likely to face interruptions, while those in Class 2 having relatively stable living conditions, not experiencing occupational mobility, and a moderate income, still face disruptions in coverage. This phenomenon highlights that factors beyond occupational income and stability, some unrecognized barriers we are not found may influence insurance participation. Low-income PWNEFs may maintain higher continuity by employing coping strategies, such as choosing low-coverage insurance scheme. Labor-intensive workers, like delivery drivers, are more likely to maintain coverage, whereas those in creative occupations, such as online streaming, face higher risks of discontinuity. These findings provide insights into the heterogeneity of CPBMI-OC among PWNFEs and emphasize the need for targeted strategies to address vulnerabilities and structural inequities of PWNFEs.

The results found that about 61.3% were uninsured, and 85.6% reported an interruption in coverage, indicating a systemic gap in maintaining the continuity of BMI among PWNFEs. The interrupt rate of the specific classes revealed differences (χ² = 30.68, P < 0.001). Class 1 reported a 100% interruption rate, with even Class 3 (despite its highest continuity) reaching 79.3% interruption. This reflects previous literature emphasizing low insurance participation rates among informal workers. Prior literature has highlighted this phenomenon; Dror [43] studied the reality that about 3 billion people employed in the informal sector in low- and middle-income countries (LMICs) do not have any form of insurance. Ho [31], using a cross-sectional survey, discovered that only 25.1% of workers in the informal sector have health insurance coverage. Previous studies by Thornton and Kaiser highlighted systemic barriers that hinder informal workers from maintaining CPBMI, such as employer opposition, income instability, and lack of formal contracts [44, 45]. However, our study identifies distinct results across three classes that reveal complexities beyond these traditional explanations. Despite having higher incomes and strong ties to formal employment contracts and family migration, Class 1 participants experienced a complete interruption in their CPBMI. This suggests that income and contracts alone do not guarantee continuity, because China’s basic medical insurance system operates under a principle of localized management [46], where individuals are required to enroll in the insurance of their residential region, leading to challenges for cross-regional insurance enrollment caused by frequent population movement and urbanization. In contrast, Class 2 comprised mid-income workers with stable residences and no employment-related floating. Although 53.5% had formal contracts, over 90% of this group experienced interruptions, indicating other barriers, such as limited insurance knowledge gaps about participation [47]. Meanwhile, Class 3, the largest group, typically individual mobility, faced the lowest incomes and an almost complete lack of formal contracts, which was a consistent barrier to CPBMI continuity. However, this group achieved the highest rate of CPBMI (21.7%). We found a positive association between URRBMI coverage and continuous CPBMI in Class 3, highlighting the role of government-subsidized insurance coverage in reducing coverage interruptions. PWNFEs can choose between UEBMI and URRBMI flexibility [14], but the significantly higher monthly premiums for UEBMI make it financially inaccessible for most workers without employer contributions [48]. Government-subsidized URRBMI’s lower annual premium offers a more affordable option, enabling broader participation among low-income individuals. Low-income groups, despite having a higher demand for medical insurance, usually face a greater financial burden and the highest proportion of individuals struggling to afford coverage [49]. Similarly, studies have shown that PWNFEs have higher enrollment rates in URRBMI [50], while participation in UEBMI remains below 20%. The integration of URBMI and NRCMS into URRBMI in 2016 addressed inequities by improving health equity and accessibility for workers in these new employment forms [51]. In recent years, China’s reforms have further relaxed the localized administration of insurance enrollment, allowing workers to enroll at their residence rather than their household registration. This adjustment effectively mitigated challenges arising from fragmented regional policies and satisfied the health needs of this workforce.

Our study expands the literature on the associated factors of BMI choice by focusing on continuous participation patterns. Age was not an associated factor in our multinomial logistic analysis, which contrasts with previous research suggesting that older informal workers are more likely to participate in health insurance [52]. This discrepancy may be attributed to the relatively young sample in our study, where younger workers face fewer age-related health challenges and medical service needs [53], resulting in lower overall participation rates. Females were more likely to interrupt their participation, contrasting with prior research emphasizing higher insurance coverage among women due to greater health awareness [54]. Workplace gender discrimination at work and societal expectations of family devotion may serve as potential reasons for females to prioritize household over personal health [55, 56]. Given that only 20.6% of the sample consists of female participants, potential gender-related selection bias may also have influenced the observed insurance participation trends among women. Urban hukou status and lower education were associated with a higher probability of CPBMI continuity (Class 3). Compared with rural hukou, workers with urban hukou are more likely to face higher medical costs [57], encouraging their participation in URRBMI to release the economic burden. According to previous studies [20], most platform workers residing in urban areas do not have an urban hukou, which distinguishes them fundamentally from Class 2. Despite this distinction, all three classes exhibit high rates of non-enrollment or discontinuity, reflecting a gap in insurance coverage. This group remains highly vulnerable regardless of hukou status, with limited access to public health resources [58]. Higher education was negatively associated with CPBMI, differing from previous findings that link higher education with greater insurance coverage [59]. This may reflect the unique systemic barriers faced by PWNFEs in China. Highly educated workers may be better equipped to assess the cost-effectiveness of insurance schemes. When faced with challenges in UEBMI access and the limited effectiveness of URRBMI, such as cross-regional reimbursement restrictions and compromised compensation, highly educated workers may choose to discontinue their participation [48]. There was a positive relationship between workers who didn’t have extra health insurance and higher CPBMI continuity (Class 3). This suggests that people who don’t have extra health insurance might rely on government-funded insurance programs like URRBMI to cover their medical needs. Public insurance serves as a crucial safety net for vulnerable populations, especially those unable to access or afford private coverage [30, 60]. These findings highlight the essential role of basic medical insurance in addressing coverage gaps for low-income workers in new employment forms. However, they also underscore the limitations of relying solely on public insurance to ensure comprehensive healthcare security. The negative association between platform linkage and Class 2 may reflect platform workers’ unfavorable perceptions of employers’ responsibilities for contributions. Despite their enrollment in UEBMI, these workers might still hold the belief that they lack the financial means to contribute without robust employer support. Prioritizing the simplification of the enrollment process and the creation of new enrollment statuses of PWNFEs in UEBMI may be a more effective measure [48].

Moderate SRH was positively associated with Class 3, and those who were poorer were more likely to maintain continuous participation. This observation is consistent with findings from studies on adverse selection in insurance markets [33, 61]. However, chronic conditions showed no significant association with CPBMI patterns. The potential reason is the younger samples of PWNFEs are less likely to develop chronic disease compared to older populations [62]. Our analysis revealed a chronic disease prevalence of less than 20%, significantly lower than the 53.8% reported among adults aged 18–34 years in the United States [63]. We did not identify medical service utilization as an associated factor. This may suggest that their frequency of healthcare use does not primarily drive decisions about continuous participation. On the other hand, the previous study has found that the BMI system has improved medical service utilization rates, particularly among low-income groups, contributing to more equity in access to medical services among these insurance schemes [64].

This study has several limitations. First, as a cross-sectional survey, this study cannot establish causal relationships between factors and patterns of CPBMI-OC. The use of retrospective questions to assess CPBMI, instead of prospective measurements, may lead to information bias [65]. Prospective cohort studies would be valuable in exploring causal links and better understanding the changes between patterns of CPBMI-OC and associated factors. Additionally, subsequent studies could explore other potential factors, such as regional policy variations or employer practices. Second, the sample size is relatively small, and the law proportion of female participants may limit the representativeness of gender-specific trends and explanation of results. Future studies should aim to include larger and more diverse samples to enhance the robustness and applicability.

Conclusion

We discovered that PWNFEs had a low BMI coverage rate of 38.7% and a low CPBMI-OC rate of only 14.4%. Through LCA, we identified three distinct behavioral patterns: fully interrupted, high-income, family migration class (Class 1); high interruption, mid-income, stable residence class (Class 2); and high continuity, low-income, non-contracted individual mobility class (Class 3). Risk factors for certain interrupt patterns include female gender, rural hukou, higher SRH, advanced education, supplementary insurance coverage, and platform-employment linkage. Conversely, the URRBMI plan emerged as a key factor associated with increased continuity. The flexible BMI enrollment options present notable challenges for PWNFEs: while UEBMI offers superior insurance benefits, its high premiums and lack of employer contributions make it largely inaccessible; on the other hand, URRBMI, with government-subsidized and lower annual premiums, is a more feasible option to maintain higher rates of continuous participation despite its relatively limited benefits. This underscores URRBMI’s suitability in addressing the current needs of PWNFEs and highlights the importance of optimizing the system to enhance coverage and continuity further. Reforms are urgently needed. Policymakers should prioritize simplifying cross-regional enrollment and reimbursement procedures. To clarify the labor relationship between platform employers and PWNFEs and to make joint contributions to the UEBMI, lawmakers have adopted legislation. Our findings hold important policy implications for designing targeted interventions for distinct patterns of CPBMI-OC.