Analyzing the Prevalence of Smartphone Addiction (SA) Among Preschoolers (3–6 Years) and Its Demographic and Familial Risk Factors: A Cross‐Sectional Study in Noakhali District, Bangladesh

2.1. Study Design and Setting

The study was conducted in Noakhali District, Bangladesh, which is predominantly a rural area with emerging semiurban (rurban) communities. The study targeted preschool children aged 3–6 years residing in the selected communities of the district. Since direct assessment of children′s smartphone use was not feasible, information was obtained from their parents or primary caregivers, who served as proxy respondents. Caregivers were included because they are the primary observers of children′s daily behaviors and smartphone use patterns, and thus can provide reliable information regarding the child′s usage.

2.2. Sampling Strategy

The study employed a multistage sampling design. In the first stage, Upazilas were selected using random sampling. In the final stage, eligible households were identified within selected clusters. However, a complete sampling frame of households with preschool‐aged children was not available, which precluded probability‐based sampling at the household level. An initial list of households was obtained from administrative records, but it did not permit systematic random selection of eligible households. Therefore, households were identified through field screening and recruited using a convenience sampling approach. This approach may have introduced selection bias and limits the representativeness of the study population. Accordingly, findings should be interpreted with caution regarding their generalizability beyond similar rural‐dominant settings.

2.3. Sample Size and Data Collection Procedure

The required sample size for the study was determined based on Roscoe (1975) guideline [23], which suggests that a sample size between 30 and 500 is generally appropriate for most behavioral research. Accordingly, a sample within this range was considered adequate for the present study.

Prior to data collection, participants were informed about the purpose of the study and invited to participate voluntarily. Written or verbal informed consent was obtained from all participants, and they were assured of the confidentiality and anonymity of their responses.

Primary data were collected by four trained undergraduate students. The data collectors underwent a structured training session conducted by the authors to ensure standardization of data collection procedures and minimize interviewer bias. Following training, interviewer‐administered face‐to‐face interviews were conducted between May 10, 2023, and June 25, 2023.

A total of 470 households were approached during the field survey. Among them, 5 households were excluded because they did not own a smartphone, and 29 households were excluded because of the absence of preschool children. Additionally, 6 eligible participants declined to participate. Therefore, the final analytical sample consisted of 441 participants, yielding a response rate of over 95%.

Data were collected using a structured, interviewer‐administered questionnaire to ensure clarity of understanding and to minimize missing or incomplete responses. The questionnaire was developed based on previously published literature on SA and PSU. Content validity was ensured through expert review by public health and statistics professionals, who evaluated and refined the instrument based on their feedback.

Before the main survey, a pilot study was conducted among 30 parents from a neighboring community who were not included in the final sample. The pilot test was used to assess clarity, comprehensibility, and feasibility of the questionnaire, and necessary modifications were made accordingly.

2.4. Study Variables and Measurements

SA among preschool children was the primary outcome variable of this study. It was assessed using an eight‐item dichotomous (yes/no) questionnaire adapted from Leung (2008) [24] and subsequently used in Bangladeshi studies by Mahmud et al. (2020) [25] and Hosen et al. (2021) [26].

Each item was scored as 1 for “yes” and 0 for “no,” resulting in a total score ranging from 0 to 8. Based on established cut‐off criteria, a score of 0–1 was classified as nonaddicted, whereas a score of 2–8 was classified as smartphone addicted.

The independent variables included a range of sociodemographic, familial, child‐related, and behavioral characteristics. These variables are presented in Table 1. Sociodemographic variables included family type, area of residence, maternal age, parental education levels, dual‐income family status, parental occupation, and monthly family income. Child‐related variables included child age, gender, number of siblings, and whether the child owned a personal smartphone.

Behavioral and parental variables included parents′ perception and behavior toward smartphone (PPBTS) use, daily smartphone use by the mother (PDMUS), and daily smartphone use by the child (PDCUS). These variables were categorized based on previous literature and conceptual relevance to smartphone use behavior.

Residential status was classified as rural and semiurban (rurban) based on administrative boundaries. Households located within municipal areas were classified as semiurban, whereas those located outside municipal areas were considered rural, consistent with the study′s focus on predominantly rural communities in Noakhali District.

In addition, residential status was classified as semiurban (rurban) and rural. Households located within municipal areas were classified as rurban, whereas those located outside municipal areas were classified as rural, to ensure alignment with the study objectives and the variables used in the analysis.

2.5. Questionnaire Development, Validity, and Reliability

The questionnaire was adapted from previously published studies on SA and PSU among young children. To establish content validity, the draft questionnaire was reviewed by experts in public health and statistics, and revisions were made based on their recommendations. A pilot survey involving 30 parents from a neighboring community who were not included in the final study was conducted to assess the clarity, comprehensibility, and feasibility of the questionnaire. Necessary modifications were made before the main survey.

The Smartphone Addiction Scale (SAS) consisted of eight dichotomous (yes/no) items. Internal consistency reliability was evaluated using Cronbach′s alpha. Construct validity was assessed using confirmatory factor analysis (CFA) with the weighted least squares mean and variance adjusted (WLSMV) estimator, which is appropriate for dichotomous variables. Model fit was evaluated using the chi‐square statistic (χ2), Comparative Fit Index (CFI), Tucker–Lewis Index (TLI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR). Cronbach′s alpha values of ≥ 0.70 were considered indicative of acceptable internal consistency. Standardized factor loadings were examined to evaluate item performance, with values of approximately ≥ 0.40 generally considered desirable; however, items with lower loadings were retained when supported by their conceptual importance and role in the predefined scoring framework.

2.6. Conceptual Framework of the Study

The conceptual framework (Figure 1) was developed based on previous literature and the study objectives to illustrate the hypothesized determinants of SA among preschoolers. It proposes that sociodemographic and familial characteristics, including parental education, occupation, income, smartphone ownership, PPBTS, and parental smartphone use, influence children′s daily smartphone use, which subsequently affects the risk of SA. Guided by this framework, the associations between these factors and SA were examined using chi‐square tests, CFA, correlation analysis, and multivariable binary logistic regression.

2.7. Statistical Analysis

Descriptive statistics were used to summarize the characteristics of the study participants and study variables. Frequency distributions and percentages were computed for categorical variables.

In the bivariate analysis, the association between SA and each independent variable was examined using the chi‐square test. Variables with a statistically significant association (p < 0.05) were considered for further analysis.

A multivariable binary logistic regression model was then performed to identify factors independently associated with SA while controlling for other covariates. Adjusted odds ratios (AORs) with 95% confidence intervals (CIs) were reported to measure the strength of association between predictors and SA.

Confounding was addressed by including variables with a p − value < 0.05 in the bivariate analysis together with variables identified a priori from previous literature as potential confounders. These variables included child age, child sex, parental education, parental occupation, family monthly income, child smartphone ownership, parental daily smartphone use, and PPBTS. A multivariable binary logistic regression model was then fitted to estimate AORs while controlling for these potential confounders. Although several important confounders were adjusted for, residual confounding cannot be completely excluded because of the observational cross‐sectional design of the study.

Multicollinearity among independent variables was assessed using tolerance values and the variance inflation factor (VIF). Tolerance values greater than 0.20 and VIF values less than 5 were considered indicative of no significant multicollinearity.

Interaction effects between independent variables were not formally assessed because of model parsimony considerations and sample size limitations.

All statistical analyses were performed using the Statistical Package for Social Sciences (IBM SPSS, Version 25.0) and R statistical software (version 4.1.2). A two‐sided p value of < 0.05 was considered statistically significance.

3.1. Reliability and Construct Validity of the SA Scale

The internal consistency (Table 2) of the eight‐item SAS was assessed using Cronbach′s alpha. The scale demonstrated good internal consistency (Cronbach′s α = 0.82).

Corrected item‐total correlations ranged from 0.11 to 0.80 (Table 3). Although Item 1 showed a relatively low corrected item‐total correlation (0.11), it was retained to preserve the conceptual coverage and predefined scoring system of the instrument. Deleting Item 1 increased Cronbach′s alpha to 0.85; however, the overall reliability of the complete 8‐item scale remained acceptable.

Construct validity of the SAS was evaluated using CFA. The hypothesized one‐factor model demonstrated excellent fit to the observed data (χ ² = 36.90, df = 20, χ ²/df = 1.85, CFI = 0.997, TLI = 0.996, RMSEA = 0.046, and SRMR = 0.069). All fit indices satisfied the recommended criteria, indicating that the proposed measurement model adequately represented the underlying construct (Table 4).

Construct validity was evaluated using CFA with the WLSMV estimator (Table 5). The one‐factor model demonstrated good model fit (χ ² = 36.90, df = 20, χ ²/df = 1.85, CFI = 0.997, TLI = 0.996, RMSEA = 0.046, SRMR = 0.069). Standardized factor loadings ranged from 0.176 to 0.999. Although Item 1 exhibited a relatively low loading (0.176), the remaining items showed moderate to very strong loadings (0.633–0.999), supporting the unidimensional structure of the SAS.

3.2. Sociodemographic and Familial Characteristics

Table 6 shows that among the 441 respondents, 50.6% are male preschoolers and 49.4% are female preschoolers. About 73% of Childs′ mothers are housewives, whereas most of the fathers (42%) are in jobs/services. 54.6% of households live in rurban areas, and 45.4% in rural areas.

Most children (85.9%) had no siblings. More than half of the mothers (56.9%) used smartphones for ≤ 3 h per day, whereas 61.9% of children used smartphones for ≥ 1 h per day. In terms of parental attitudes, the largest proportion of parents (38.5%) reported positive perceptions and behaviors toward smartphone use, whereas 32.3% reported negative perceptions.

Overall, 69.0% of preschool children were classified as having SA (Figure 2), with a slightly higher proportion among males (52.4%) compared with females (47.6%).

3.3. Association Between SA and Independent Variables

Table 7 presents the results of the chi‐square tests examining the association between SA and selected sociodemographic and familial variables. Significant associations with SA were observed for family type, maternal age, dual‐income family status, father′s occupation, ownership of a personal smartphone, number of siblings, PPBTS, daily maternal smartphone use (PDMUS), and daily child smartphone use (PDCUS) (p < 0.05).

No statistically significant associations were observed for residential status, child age, child gender, maternal education, father′s education, or family monthly income (p > 0.05).

3.4. Multicollinearity Diagnostics

Multicollinearity among the independent variables was assessed using tolerance and variance VIF. Tolerance values ranged from 0.384 to 0.865, whereas VIF values ranged from 1.156 to 2.605 (Table 8). All VIF values were below three, indicating no evidence of problematic multicollinearity among the predictor variables.

3.5. Multivariable Logistic Regression Analysis

Table 9 shows that, after adjusting for potential confounders, father′s occupation, child ownership of a smartphone, parental perception and behavior toward smartphones, and daily child smartphone use remained statistically significant predictors of SA.

Children whose fathers were engaged in service or job‐related occupations had significantly lower odds of SA compared with those whose fathers were involved in business (AOR = 0.27, 95% CI: 0.123–0.592). Children who did not own a smartphone had significantly lower odds of SA compared with those who owned a personal device (AOR = 0.079, 95% CI: 0.027–0.230).

Parental perceptions and behaviors regarding smartphone use were strongly associated with SA. Compared with children whose parents reported more positive attitudes, those with negative perceptions had significantly lower odds of SA (AOR = 0.069, 95% CI: 0.026–0.181).

Children who used smartphones for more than 1 h per day were more likely to have SA compared to those using smartphones for less than 1 h per day (AOR = 5.416, 95% CI: 2.800–10.476).

Table 10 presents the model diagnostics for the logistic regression analysis. The Hosmer–Lemeshow goodness‐of‐fit test indicated an acceptable model fit (χ ², d f = 8, p = 0.725). The Nagelkerke R‐squared value of 0.587 suggests that approximately 58.6% of the variation in SA was explained by the variables included in the model.