Socioeconomic and environmental disparities in childhood unintentional injury risk in urban households in Chongqing, China: a latent profile analysis of cross-sectional survey data

Abstract

Objectives

While the social determinants of the home environment and childhood unintentional injuries (UIs) are well established, their association within the context of rapid urbanisation in China remains understudied. Given its rapid urban growth and unique housing landscape, Chongqing provides a valuable context for this investigation. This study aimed to investigate how socioeconomic and home environmental factors collectively shape the risk of childhood UIs in urban households in Chongqing.

Design

Cross-sectional study.

Setting

Data were collected in April 2023, in Jiangbei District, Chongqing Municipality, China.

Participants

A total of 1282 parents of children aged 0–14 years were recruited by stratified convenience sampling by the local Center for Disease Control and Prevention. Ultimately, 1186 valid questionnaires were included in the data analysis.

Outcome measures

Various indicators, including family demographics, home environmental characteristics and in-home environmental UI risk assessment, were examined.

Results

Three risk levels were identified by latent profile analysis: the Secure Environment Profile (70.91%), the Intermediate-Risk Profile (24.20%) and the High-risk Multi-hazard Profile (4.89%). Key determinants were identified through ordinal logistic regression. An increase in living area per child (m²) was associated with a reduced likelihood of being classified within a higher risk profile (OR=0.995, 95% CI 0.990 to 1.000). Furthermore, compared with households with a monthly per capita income of ≥15 000 yuan, those earning <4000 yuan and 4000–6999 yuan had 2.004 (95% CI 1.177 to 3.408) and 1.893 (95% CI 1.168 to 3.068) times the odds, respectively, of being classified within a higher-risk profile.

Conclusions

This study identifies significant heterogeneity in childhood UI risk among urban households in Chongqing, effectively characterised by distinct risk profiles. Living area per child (m²) and monthly per capita income were identified as the major social determinants, providing a basis for developing targeted prevention and intervention measures for vulnerable urban children in China.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Latent profile analysis was employed as a data-driven method to capture hidden risk patterns within the study population.

• The ordinal logistic regression modelling incorporated a combined analysis of individual demographic, socioeconomic and home environmental characteristics.

• The reliance on caregiver-reported questionnaires may introduce recall and reporting biases due to the subjective nature of the responses.

• The stratified convenience sampling strategy could lead to selection bias, limiting the generalisability of findings.

• The cross-sectional design restricted the ability to establish causal relationships between the identified determinants and childhood unintentional injuries.

Introduction

Unintentional injuries (UIs) remain the leading cause of child mortality and morbidity worldwide. In China, data from the National Injury Surveillance System (NISS) indicate that the home is the primary location for UIs among both children and adolescents.^{1 2} The spaces where children grow up, including housing, streets, neighbourhoods and cities, are a reflection of underlying socioecological patterns.³ However, socioeconomic and environmental disparities within urban environments, including disparities in service distribution, high living costs and poor built environments, disproportionately affect vulnerable families and likely contribute to the injury risks observed at the household level.^{4 5}

While many studies have described the socioeconomic factors associated with childhood UIs in the household,^{6 7} there remains limited evidence regarding the effectiveness of specific environmental interventions.⁸ This lack of interventional success suggests that the specific role of the physical home environment, and particularly how it interacts with socioeconomic factors to shape injury risks, remains insufficiently explored. In the context of China’s rapid urbanisation, residential environments have undergone profound transformations, characterised by rising population densities, a transition from traditional multistorey buildings to modern high rises, an increasingly diverse array of architectural layouts and varying living standards.⁹ These changes have introduced indoor safety challenges for children, such as the persistently high fall risks over recent years from unsecured windows and balconies in high-rise buildings, or hazardous spatial layouts in older residential areas.¹⁰ However, research analysing how these structural layouts of urban homes in China influence childhood injuries remains scarce.¹¹ Even fewer studies have integrated socioeconomic and environmental risk factors, making their collective impact on childhood UIs poorly understood.¹²

Located at the crossroads of the Yangtze River Economic Belt, Chongqing is the only direct-administered municipality in central and western China. As one of the most rapidly urbanising cities in the country, it had a resident population of 32.13 million by the end of 2022, including 4.56 million children aged 0–14 years, with 22.80 million residing in urban areas.¹³ The city’s unique socioeconomic context, characterised by rural–urban migration and income disparities, further influences injury risks by creating uneven access to safety resources and awareness. Empirical research from similar rapidly urbanising settings like Shenzhen City, which surveyed elevator apartments, stair-access buildings and urban villages, reveals that safety interventions should primarily target families living in urban villages.¹⁴ In these areas, risks such as unprotected electrical outlets and unshielded sharp furniture corners were prevalent. Beyond these vulnerabilities, behavioural factors such as limited housing safety awareness and inconsistent parental supervision further compound injury risks.

In this study, socioeconomic factors are defined as household-level indicators, including caregiver education and household income, which may influence a family’s resources and ability to provide a safe environment and effective supervision.^{6 7} Housing environmental factors are specifically defined as the microlevel physical characteristics of the home, such as structural layout (eg, floor level, housing form), interior conditions (eg, flooring materials, lighting) and safety facilities (eg, window guards, electrical outlet covers). These definitions exclude macrolevel neighbourhood or community characteristics, focusing instead on the immediate physical spaces where children interact with their environment.

Latent profile analysis (LPA), a person-centred analytical approach, is used in this study to classify children and families into subgroups based on their shared risk for childhood UIs. Although traditional risk classifications often rely on predefined cut-offs for low, medium and high risk, LPA offers greater flexibility by using probabilistic modelling to refine these classifications. Its accuracy, objectivity and ability to handle complex heterogeneity make it a useful tool for identifying subgroups that share similar injury risk profiles, which can subsequently guide more tailored and effective interventions.¹⁵

By integrating individual, socioeconomic and environmental factors, this study aims to clarify underlying determinants of childhood UIs and provide evidence for designing targeted interventions that address the specific safety needs of diverse subgroups in the rapidly urbanising context in Chongqing.

Methods

Design

This was a cross-sectional, descriptive study that was conducted in April 2023 in Chongqing, China. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) and the Checklist for Reporting Of Survey Studies (CROSS) checklists were used to guide the report of this study.

Participants

The sampling frame for this study comprised parents of children aged 0–14 years residing in Jiangbei District, Chongqing Municipality. Families were selected from the Lanxi and Haiyu Jiangshan communities, which represent the central developed areas and relatively less-developed peripheral areas of the district, ensuring coverage of diverse residential and socioeconomic backgrounds. The inclusion criteria were: (1) having at least one child aged 0–14 years, (2) being the primary caregiver of the child and (3) being reachable by smartphone and having reliable internet access to complete the study’s procedures. The exclusion criteria were: (1) caregivers of children with known chronic illnesses or disabilities affecting mobility or daily activities and (2) individuals unwilling or unable to provide informed consent.

An a priori power analysis using G*Power V.3.1 indicated a minimum requirement of 191 participants for a regression model with 22 predictors (α=0.05, power=0.90, R²=0.15). After adjusting for a 30% non-response and missing data rate, the calculated sample size was 273.

Recruitment

Participants in this study were recruited via the Chongqing Jiangbei Center for Disease Control and Prevention. A stratified convenience sampling approach was employed to ensure diversity across key characteristics. The stratification was based on two factors: (1) type of community within the urban district (eg, central developed areas and relatively less-developed peripheral areas) and (2) housing environment (eg, older multistorey apartments and modern high-rises). Investigators collaborated with community leaders and school staff, who used demographic records to identify and invited 1485 potentially eligible families. Of these, 1307 families responded to the initial invitation and expressed interest in participating. These families received a QR code linking to an electronic survey platform (wjx.cn) to access the questionnaire, which they were asked to complete within 2 weeks. On scanning the code, participants first completed a preconsent eligibility checklist. Eligible individuals were automatically directed to the informed consent form detailing the study’s purpose and requirements. To ensure data quality, all questionnaire items were set as mandatory to prevent missing data, and a WeChat login was required to restrict submissions to one per account. Ultimately, 1282 participants successfully completed the screening, consent process and the final questionnaire (meaning 25 individuals were either deemed ineligible during the online check or dropped out before submission). Three researchers independently reviewed the dataset to verify data accuracy.

Measures

The questionnaire used in this study was provided in online supplemental file 1.

Demographic information

Demographics including the child’s age, gender, relationship to the primary caregiver, number of children and permanent residents in the household, educational levels of the child’s mother, father and/or the primary caregiver, family structure and the household income were self-reported.

The Environmental Scale of Unintentional Injury in the Home for Children Aged 0–6 Years in Urban Areas of China (abbr. the In-home Environmental Risks Scale)

The scale was developed by Wang et al.¹⁶ The first section of the scale was 6 items on the physical environment of the home, including the housing tenure type, housing form and house size as well as the type of flooring in the bedrooms, living room and kitchen and bathroom. The second section contained 54 items on the assessment of six types of UIs inside the children’s homes, including: falls (15 items), external force injuries (12 items), burns (12 items), poisoning (7 items), foreign body injuries (6 items) and animal-related injuries (2 items). Respondents answered on a 5-point scale ranging from 1 (fully compliant), 2 (mostly compliant), 3 (partially compliant), 4 (mostly non-compliant) to 5 (completely non-compliant). If an item did not match the home setup, respondents selected ‘not applicable’, which was scored as 1. The scale is designed to evaluate home environment safety for UI prevention among children in urban areas, with higher scores indicating greater in-home risk. Based on the scale’s cut-off values, family environments were classified into low risk (<71 points), moderate risk (71–88 points) and high risk (≥89 points).¹⁶ Consent to use the scale in this study was obtained from the authors. Although risks from schools and public areas become more prominent as children grow older, risk factors within the home remain consistent. Based on this contextual consistency, the scale remained appropriate for assessing home injury risks across the broader age range of 0–14 years in our study. Furthermore, the Cronbach’s alpha coefficient of the scale was 0.943 in this study, indicating excellent internal reliability when applied to this wider age group.

Data analysis

SPSS V.23.0 and Mplus V.8.7 were used for data analysis. Demographic characteristics were described using frequency counts and proportions. The total scores of the six dimensions of the scale were standardised to a 5-point scale for LPA. The LPA began with a single-profile model and subsequently progressed to the estimation and comparison of models with increasing numbers of classes. The optimal class solution was identified based on the following fit indices: (1) information criteria including Akaike’s Information Criterion (AIC), Bayesian Information Criterion (BIC) and sample size-adjusted BIC (aBIC), where lower values indicating improved model fit; (2) entropy values approaching 1, with values exceeding 0.80, indicating classification accuracy above 90%; (3) likelihood ratio tests such as the Lo-Mendell-Rubin-adjusted likelihood ratio test (LMRT) and the Bootstrap Likelihood Ratio Test (BLRT), with significance levels (p<0.001) supporting the superior fit of the k-class model over the k-1 class model and (4) adequate class size, with each class representing more than 5% of the total sample.¹⁷

Given the non-normality of the data, the Kruskal-Wallis rank sum test was employed to assess differences in the scores of the six injury types and the total scale score across the identified latent profiles. Additionally, an ordinal logistic regression was conducted using children’s and family demographic and environmental characteristics to identify predictive factors for the injury risk. P<0.05 was considered to be statistically significant.

Patient and public involvement

None.

Results

Demographics of the study sample

The study recruited 1282 parents of children aged 0–14 years. Responses were excluded if the questionnaire was completed in under 5 min, if the child was older than 14 years, or if there were inconsistencies between parent and child information. A total of 1186 valid questionnaires were included in the data analysis, with an effective response rate of 79.87%. The gender distribution of the children was almost equal. The majority of families were nuclear and core families. The number of members in the household ranged from 2 to 8, with 3 to 6 members being the most common. In most cases, parents were the primary caregivers of the children, followed by grandparents. Furthermore, 39.04% of respondents had a bachelor’s degree or higher. Regarding economic status, 16.27% of households had an average monthly per capita household income of less than 4000 yuan, while 28.75% had an income of ≥10 000 yuan (table 1).

Table 1. Demographic characteristics of children and their caregivers (N=1186).

Category	n	%
Number of boys
0	479	40.39
1	599	50.51
2	104	8.77
3	4	0.33
Number of girls
0	500	42.16
1	585	49.33
2	95	8.01
3	5	0.42
4	1	0.08
Child’s gender
Male	584	49.24
Female	602	50.76
Child’s age (year, median: 8)
1-3	92	7.76
4-6	310	26.14
7-14	784	66.10
Family structure*
Nuclear family	604	50.93
Core family	462	38.95
United family	34	2.87
Single-parent family	45	3.79
Reorganised family	17	1.43
Intergenerational family	24	2.03
Number of members in the household
2	40	3.37
3	283	23.86
4	399	33.64
5	292	24.62
6	158	13.32
7	8	0.67
8	6	0.52
Primary caregivers of children†
Parent(s)	1013	85.41
Grandparent(s)	382	32.21
Other family member(s)	2	0.17
Childcare services	3	0.25
Educational attainment of primary caregiver
Junior high school or below	381	32.12
Senior high school	342	28.84
College	454	38.28
Master or above	9	0.76
Monthly household income per capita (yuan)
＜4000	193	16.27
4000–6999	379	31.96
7000–9999	273	23.02
10000–14999	209	17.62
≥15 000	132	11.13

^*The composition of family structures is: (1) Nuclear family: parents and children. (2) Core family: grandparents, parents, and children. (3) United family: multiple generations including grandparents, parents, children, and grandchildren. (4) Single-parent family: one parent with children. (5) Reorganised family: single parents with children after remarriage. (6) Intergenerational family: grandparents and children.

^†Multiple-choice question with no fixed terms.

Home environmental characteristics and risks

The surveyed households had living areas ranging from 40 to 300 m², with a median of 90.00 m². In total, 88.20% of the households were home owners. A majority of households lived in high-rise buildings (92.50%). Residential floors were distributed from the 1st to the 40th floor, with a median of the 9th floor (online supplemental table 1).

The overall score on the In-home Environmental Risks Scale was 64 (IQR: 57–87). Using the original stratification criteria, the low risk (<71), moderate risk (71–88) and high risk (≥89) groups accounted for 59.11%, 17.03% and 23.86% of the sample, respectively.¹⁶ Overall, falls showed the highest risk levels (online supplemental table 2).

Latent profile analysis

Model fit indices for the one-profile to five-profile models were evaluated, indicating a continuous decrease in AIC, BIC and aBIC values as the number of profiles increased. Entropy values surpassed 0.8 in all models, with the peak entropy observed in the 3-profile model. P values of the LMRT and BLRT for the two-profile to four-profile models were below 0.001, whereas the p value for LMRT in the five-profile model exceeded 0.05, supporting the adoption of the four-profile model. Additionally, in the four-profile model, classes 2 and 3 exhibited similar trends regarding in-home UI risk with minimal score differences (table 2). Taking these results into account, the three-profile model was selected as the optimal solution.

Table 2. Summary of latent profile analysis fitting information (N=1186).

Model	AIC	BIC	aBIC	Entropy	P		Categorical probability (%)
					LMRT	BLRT
1	11 649.997	11 710.937	11 672.821	–	–	–	1.00
2	7205.340	7301.829	7241.478	0.970	<0.0001*	<0.0001*	76.31/23.69
3	5518.101	5650.138	5567.552	0.976	<0.0001*	<0.0001*	70.91/24.20/4.89
4	4952.054	5119.639	5014.819	0.950	0.0009*	<0.0001*	62.22/16.61/16.70/4.47
5	4640.237	4843.370	4716.316	0.955	0.4490	<0.0001*	63.41/15.68/2.95/3.29/14.67

^*P<0.001.

aBIC, adjusted BIC; AIC, Akaike’s Information Criterion; BIC, Bayesian Information Criterion; BLRT, Bootstrap Likelihood Ratio Test; LMRT, Lo-Mendell-Rubin-adjusted likelihood ratio test.

The three-profile model showed significant differences across the six dimensions of in-home injury risks (figure 1). Class 1 (70.91%, n=841) scored the lowest across all six dimensions and was thus named the Secure Environment Profile. Class 2 (24.20%, n=287) had median scores with a slightly higher risk of foreign body injuries and was labelled the Intermediate-Risk Profile. Class 3 (4.89%, n=58) scored the highest, indicating the greatest risk of foreign body injuries, external force injuries and poisoning, earning it the High-risk Multi-hazard Profile. Kruskal-Wallis tests confirmed a hierarchical risk among the three risk profiles across the six dimensions and the overall scale (p<0.001).

Figure 1. Distribution of features across the three latent profiles of childhood UIs in the household. UIs, unintentional injuries.

Determinants affecting the classification of in-home risk profiles regarding childhood UIs

Using the information of children, families and living environments as independent variables, a univariate analysis using Kruskal-Wallis test was conducted across the three latent profiles. Indicators with p<0.1 in the univariate analysis (table 3: categorical variables coded as dummy variables), along with continuous variables (child age, residential floor, living area per child) were included in an ordinal logistic regression analysis to identify determinants affecting the latent profiles of children’s in-home injury risk.

Table 3. Categorical difference of latent profiles regarding in-home UI risks among children (N=1186).

	n (%)			H*	P
	Low risk	Moderate risk	High risk
Educational attainment of primary caregiver
Junior high school or below	256 (67.19)	99 (25.98)	26 (6.83)	6.356	0.096
Senior high school	243 (71.05)	85 (24.85)	14 (4.10)
College	337 (74.23)	99 (21.81)	18 (3.96)
Master or above	5 (55.56)	4 (44.44)	0 (0.00)
Monthly household income per capita (yuan)
＜4000	122 (63.21)	58 (30.05)	13 (6.74)	22.189	<0.001†
4000–6999	250 (65.96)	105 (27.70)	24 (6.34)
7000–9999	199 (72.89)	66 (24.18)	8 (2.93)
10 000–14 999	164 (78.47)	37 (17.70)	8 (3.83)
≥15 000	106 (80.30)	21 (15.91)	5 (3.79)
Flooring of bedrooms
Wood floor	507 (71.31)	172 (24.19)	32 (4.50)	8.376	0.079
Anti-slip ceramic tiles	289 (72.80)	89 (22.42)	19 (4.78)
Non-slip ceramic tiles	28 (59.57)	13 (27.66)	6 (12.77)
Cement floor	16 (57.14)	11 (39.29)	1 (3.57)
Other	1 (33.33)	2 (66.67)	0 (0.00)
Flooring of living rooms
Wood floor	95 (66.90)	37 (26.06)	10 (7.04)	11.470	0.022‡
Anti-slip ceramic tiles	619 (73.43)	189 (22.42)	35 (4.15)
Non-slip ceramic tiles	109 (64.88)	47 (27.98)	12 (7.14)
Cement floor	17 (56.67)	12 (40.00)	1 (3.33)
Other	1 (33.33)	2 (66.67)	0 (0.00)
Flooring of kitchen and bathrooms
Wood floor	10 (83.34)	1 (8.33)	1 (8.33)	15.236	0.002†
Anti-slip ceramic tiles	778 (72.30)	249 (23.14)	49 (4.56)
Non-slip ceramic tiles	45 (54.88)	30 (36.59)	7 (8.53)
Cement floor	8 (50.00)	7 (43.75)	1 (6.25)

^*Kruskal-Wallis tests.

^†P<0.01.

^‡P<0.05.

UI, unintentional injury.

The ordinal logistic regression results indicated a significant overall association between household per capita monthly income and in-home injury risk (χ²=14.809, df=4, p=0.005). Specifically, post-hoc pairwise comparisons revealed that, compared with households with a per capita monthly income of ≥15 000 yuan, those with <4000 yuan showed a twofold increase in injury risk (OR=2.004, 95% CI 1.177 to 3.408, χ²=6.562, p=0.010); households with 4000–6999 yuan also exhibited an increased risk (OR=1.893, 95% CI 1.168 to 3.068; χ²=6.707, p=0.010). Additionally, living area per child was identified as a protective factor. For every additional square metre, the odds of in-home injury risk decreased (OR=0.995, 95% CI 0.990 to 1.000; χ²=4.122, p=0.042) (table 4).

Table 4. Ordinal logistic regression modelling of in-home UI risks among children (N=1186).

	β	SE	Wald	P	Exp(β)	95% CI
						Lower	Upper
The living area per child (m2)	−0.005	0.002	4.122	0.042*	0.995	0.990	1.000
Monthly household income per capita (yuan)	–	–	14.809	0.005*	–	–	–
<4000 vs ≥15 000	0.695	0.271	6.562	0.010*	2.004	1.177	3.408
4000–6999 vs≥15 000	0.638	0.246	6.707	0.010*	1.893	1.168	3.068
7000–9999 vs≥15 000	0.303	0.260	1.363	0.243	1.354	0.814	2.255
10 000–14 999 vs≥15 000	0.069	0.278	0.062	0.804	1.071	0.621	1.848

Model fitting information: χ2=27.746, p=0.006<0.01.

^*P<0.05.

UI, unintentional injury.

Discussion

Home environment affects child health. Common causes of home injuries have been discussed since the 1990s.¹⁸ In Western countries, it is well recognised that the home environment and childhood injuries are socially distributed. However, the risk factors for childhood UIs in urban households in China have not been thoroughly investigated. One reason for this is the lack of comprehensive screening research on housing safety hazards in China. Current instruments designed to evaluate the risk of childhood UIs predominantly focus on knowledge, attitudes and behaviours, while overlooking the environmental dimension.¹¹ In contrast, Western researchers have linked UIs to indicators of housing conditions. These indicators are diversified across the age of the housing, tenure type (owner-occupied or rental), built form, inadequate or deferred maintenance, high tenant turnover, overcrowding, condensation, access to a garden and heating equipment.^{5 19–21} In a multicentre case–control study conducted by Leblanc et al, researchers used home visits to observe, identify and record risk factors and safety devices for UIs (such as falls, poisoning or ingestion, burns or scalds).²² In subsequent years, various agencies and researchers have developed specific assessment tools. For instance, the Home Housing Quality Standards (HQS) Inspection Form contains 120 items that focus on structural housing elements and specific neighbourhood features as well as a few injury prevention measures.²³ Based on the HQS, Shields et al developed the Child Housing Assessment for a Safe Environment (CHASE), which includes 25 items distributed across 12 subdomains.²⁴

In the present study, we used the In-home Environmental Risks Scale developed by Chinese researchers and published in 2020.¹⁶ This scale facilitated the assessment of residential layout, safety infrastructure and hazard mitigation measures, enabling the identification of safety hazards for children within urban domestic environments in China. The children in this study had the highest risk of falls, which is consistent with the NISS data on injuries among Chinese children aged 0–18 years.^{1 2} The mean total score of the scale was lower than the results of the survey conducted by Wang et al among families of children aged 0–6 years in Changsha City, Hunan Province in central China (injury group: 86.45±21.65; non-injury group: 81.60±21.18).¹⁶

LPA was used to identify three risk profiles of childhood UIs, accounting for 70.91%, 24.20% and 4.89% of the sample, ranging from low to high risk, respectively. In comparison, the conventional tertile method distributed the sample into risk groups at 59.11%, 17.03% and 23.86%. Unlike the tertile-based approach, which relies on rank ordering and may overestimate the proportion of high-risk participants due to cut-off values, LPA: (1) uses maximum likelihood estimation and analyses the covariance structure between observed variables, thereby capturing sample heterogeneity and identifying distinct latent subgroups across six injury categories; (2) employs posterior probability estimations for risk stratification, offering superior discriminant validity over traditional stratification approaches; (3) evaluates a series of models with varying numbers of profiles using fit indices such as AIC, BIC and entropy.^{15 25} This data-driven approach facilitated a more accurate identification of children at high risk for in-home injuries.

Specific individual and family characteristics, particularly younger age, non-nuclear family structures and myopia, have been identified as key determinants associated with injuries.^{6 26 27} A study in the USA found that housing characteristics (tenure and build type) served as the significant mediators between social conditions and childhood injuries.²¹ Similarly, the UK Millennium Cohort Study aimed to examine the housing environment as a potential mediator between socioeconomic circumstances and childhood injuries.⁵ Proxy indicators of housing quality (such as build type, floor level, garden access and rooms per capita, etc) and safety product use (eg, fire guards and safety gates) were analysed as the explanatory variables; however, these indicators did not fully explain the socioeconomic inequalities in injuries.

In this study, an increased residential area per child (m²), coupled with higher per capita monthly household income, was associated with a reduced likelihood of being classed into a higher risk profile (OR=0.995, 95% CI 0.990 to 1.000 per 1 m² increase). While adding a single square metre barely changed the risk, the practical impact becomes evident when scaled up. A 10 m² increase per child yields a cumulative OR of 0.951 (0.995¹⁰). This roughly 4.9% drop in risk demonstrates that adequate living space offers tangible protection against in-home injuries. Economically disadvantaged households may neglect to address home safety hazards due to the competing time and financial priorities for safety features. Conversely, households with higher incomes may have sufficient budgets for housing improvements and access to home safety products, making them more effective in ensuring the safety of the home’s physical environment. Furthermore, larger living spaces facilitate functional zoning, reducing the likelihood of UIs such as falls and external force injuries. These factors collectively contribute to the social distribution of physical environmental risks in home settings.

Children in the high-risk and moderate-risk profiles may be key target groups for interventions against common types of home hazards. There is a need for housing surveillance and the training of public health and housing professionals in areas such as home visits, housing inspections and parental health education. Programmes that conduct safety inspections in children’s homes should incorporate home safety checklists as part of their inspection protocols. Safe Kids USA provides a series of safety checklists (eg, home safety, medication safety and fire safety) and tips tailored to different child ages and risk types.²⁸ The University of Nottingham’s Keeping Children Safe at Home (KCS) project also recommends the use of a home safety checklist to prevent accidents and injuries.²⁹ Current research on child home safety assessments tailored for urban China is limited, indicating a need for further investigation. Implementing localised safety checklist is useful for identifying and eliminating risk factors in the household to create a safer environment for children in urban households in China. Crucially, local cultural and socioecological contexts need to be taken into account when developing such checklists.³⁰ For example, residential buildings in Chongqing are mostly high-rise buildings, and bedrooms are usually designed with bay windows. This architectural feature increases the risk of children climbing up and sustaining fall injuries or even death, making the installation of window locks or restrictors necessary.

The main strength of this study was the combined analysis of individual demographic, socioeconomic and home environmental characteristics. This study used LPA, a data-driven approach that identifies subgroups within the sample. Unlike traditional methods relying on cut-offs, LPA captures hidden risk patterns within the data more effectively, improving analytical precision and supporting targeted prevention strategies. However, there were several limitations. First, the use of caregiver-reported questionnaires may introduce recall bias, as responses rely on memory and subjective interpretation. Second, selection bias was a potential limitation due to the convenience-based recruitment strategy, which may reduce the sample’s representativeness and generalisability. Third, the cross-sectional study design limited the ability to identify causal relationships between socioeconomic conditions, housing environments and UIs in children. Finally, while the study reported an effective response rate of 79.87%, the potential for non-response bias remains. Families who did not participate, or whose responses were excluded due to invalidity, may differ from the included participants regarding their living conditions or injury risks. Future studies should incorporate longitudinal designs and expand to include additional urban areas in China to improve generalisability and better understand the temporal patterns of childhood UIs.

Conclusion

This study used LPA to identify three profiles of household injury risk among urban children, classified as the Secure Environment Profile, the Intermediate-Risk Profile, and the High-risk Multi-hazard Profile. Living area per child (m²) and monthly per capita income were identified as the major predictive factors determining risk stratification.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.