Abstract

Background

Watching short videos is an integral part of the daily lives of young and middle-aged people. Nevertheless, the correlation between the screen time spent watching short videos at bedtime and essential hypertension in young and middle-aged people remains unclear. We aimed to explore the correlation between the screen time spent watching short videos at bedtime and essential hypertension among young and middle-aged people and construct a nomogram prediction model for assessing the probability of developing essential hypertension for these age groups.

Methods

This study included 4318 young and middle-aged people who underwent medical examinations at Hengshui People’s Hospital between January 2023 and September 2023. The collected data, including self-reported screen time spent watching short videos at bedtime and general information, were partitioned into a training set and a test set, with the former being divided into hypertensive and non-hypertensive groups. R programming language was used for statistical analysis and processing.

Results

The results of multifactorial logistic analysis showed that screen time of 0< time ≤ 1Ìýh (95% confidence interval [CI]: 2.022–6.082, P<0.05), 2< time ≤ 3Ìýh (95% CI: 1.538–4.665, P<0.05), 3< time ≤ 4Ìýh (95% CI: 5.327–16.691, P<0.05), and time>4Ìýh (95% CI: 21.382–78.15, P<0.05) were independently associated with essential hypertension among young and middle-aged people. Sex, age, screen time, occupation, high-sodium diet, physical activity, sleep, overweight or obesity, diabetes or glucose tolerance abnormality, dyslipidaemia, hyperuricaemia, and family history of hypertension were screened to construct a nomogram prediction model. The model had an area under the curve of the participant’s work characteristics of 0.934 (95% CI: 0.925–0.943), along with a preferably fitted calibration curve. After model validation using the test dataset, the area under the working characteristic curve for participants was 0.911 (95% CI: 0.895–0.928), and it was a well-fitted calibration plot.

Conclusions

The screen time spent watching short videos at bedtime was significantly associated with essential hypertension in young and middle-aged people, and the nomogram was a good predictor of the risk of essential hypertension among young and middle-aged people.

Introduction

Essential hypertension can cause various cardiovascular complications and seriously jeopardise human health. An increasing prevalence of essential hypertension persists among young and middle-aged people, suggesting an increased risk of cardiovascular complications [1, 2]. A poor lifestyle is an important factor associated with the essential hypertension onset among young and middle-aged individuals [3]. Identifying a poor lifestyle and correcting it are crucial for effectively controlling blood pressure and improving patient prognosis. With the rapid development of the short video industry, watching short videos is a new lifestyle for people, and studies have shown that the screen time spent on television and online games is a risk factor for cardiovascular complications [4]. However, research concerning the association of screen time with hypertension in adolescent populations have presented inconsistent conclusions [5, 6]. Moreover, the correlation of the screen time spent watching short videos at bedtime with hypertension among young and middle-aged populations has scarcely been reported, and comprehensive and effective hypertension prevention and treatment strategies for these age groups are lacking. Therefore, this study aimed to determine whether there is a correlation between the screen time spent watching short videos at bedtime and essential hypertension among young and middle-aged people and construct a nomogram-based predictive model for essential hypertension for these age groups.

Methods

Participants

Data of 4318 young and middle-aged adults who underwent medical checkups in Hengshui People’s Hospital from January 2023 to September 2023 were consecutively collected for a cross-sectional study, including self-reported screen time spent watching short videos at bedtime and general information. Initially, an overall pool of 7,246 participants were screened. From this pool, 2,112 individuals were excluded based on age criteria, with 622 participants under 18 years old and 1,490 participants over 59 years old. An additional 5 participants were excluded due to a diagnosis of secondary hypertension. A further 386 individuals were excluded for heart diseases, including 373 with coronary heart disease, 4 with congenital heart disease, and 9 with dilated cardiomyopathy. Moreover, 399 participants were excluded due to other underlying conditions, comprising 71 with chronic obstructive pulmonary disease, 7 with peptic ulcers, 14 with chronic renal failure, 286 with old cerebral infarction, and 21 with old cerebral hemorrhage. Twenty-two individuals were excluded because of malignant tumors, including 18 with lung cancer, 3 with liver cancer, and 1 with prostate cancer. Lastly, 4 participants were excluded due to psychiatric disorders, including 1 with schizophrenia and 3 with bipolar disorder. This rigorous screening process resulted in a final sample size of 4,318 participants(Fig.Ìý1).The general information collected included demographic factors (e.g., age, sex, occupation, educational background), lifestyle factors (e.g., smoking status, alcohol consumption, physical activity levels, dietary habits such as high sodium intake), and other relevant variables that may influence health. Among the 4318 participants, 2425 were male and 1893 were female, with a mean age of 32.77 ± 10.22 years. The participants included civil servants, professionals, technicians, and workers who typically undergo routine annual health checkups, as well as individuals who underwent pre-employment or pre-admission health screenings and others who voluntarily participated in health checkups. These screenings typically included basic health assessments, such as blood pressure measurements, blood tests, and other general health evaluations.The datasets were divided into training (from 3023 participants from January 2023 to July 2023) and test (from 1295 participants from July to September 2023) sets in a ratio of approximately 3:1 based on the chronological order of participation in the medical checkups. The training set was divided into hypertensive (n = 888) and non-hypertensive (n = 2135) groups depending on whether comorbid essential hypertension was present or not. Inclusion criteria were (1) 18–59 years of age; (2) hypertension diagnosis according to the diagnostic criteria of the 2018 Chinese guidelines for hypertension management [7], systolic blood pressure (BP) level ≥ 140 mmHg and/or diastolic BP level ≥ 90 mmHg; (3) no history of drug abuse; (4) habit of watching short videos at bedtime or never watching short videos both of which have lasted for > 3 months; and (5) full informed consent provided to the medical examiners. Exclusion criteria were (1) secondary hypertension; (2) heart diseases; (3) underlying diseases, such as respiratory, digestive, urinary, immune, and nervous system diseases; (4) malignant tumours; (5) severe psychiatric illnesses, including schizophrenia, bipolar disorder, and other psychiatric disorders that significantly impair an individual’s ability to perform daily activities; and (6) inability to report the time spent watching short videos at bedtime for various reasons. This work was approved by Hengshui People’s Hospital’s Ethics Committee under No: AF /SC-08/02.02.

Fig. 1

Flowchart of participant inclusion and data set division

Data collection

We collected the names, sexes, ages, and other personal information of all participants and conducted relevant medical history enquiries, physical examinations, and relevant auxiliary examinations. Using a questionnaire, we collected data on the average daily time spent watching short videos in the past 1 week, divided into 0, 0–1 (encompassing 1), 1–2 (encompassing 2), 2–3 (encompassing 3), 3–4 (encompassing 4), and > 4Ìýh. Marital status was categorised as single, married, divorced, or widowed; literacy attainment as primary school and below, junior high school, senior high school, or university and above; and occupation as unemployed, farmer, civil servant, professional and technical personnel, employee, student, or freelancer. In addition, patients were assessed on whether they smoked cigarettes, drank alcohol, were on a high-sodium diet based on a sodium intake of ≥ 6Ìýg/day [8], and were on a weekly 150Ìýmin of moderately intense or 75Ìýmin of highly intense physical activity for the past 1 week, or a mixture of both exercises at an equivalent time, defined as sufficient physical activity [9]. We also evaluated sleep adequacy based on whether the average daily sleep duration over the past week exceeded 7Ìýh [10]. Additionally, weight status was evaluated using BMI, with obesity defined as a BMI ≥ 28.0Ìýkg/m² and overweight defined as a BMI ≥ 24.0Ìýkg/m² but < 28.0Ìýkg/m² [11]. Anxiety degree was evaluated using the Self-rating Anxiety Scale comprising 20 entries, with a score of ≥ 50 categorised as having anxiety. The degree of depression was assessed using the Self-rating Depression Scale comprising 20 entries, with a score of ≥ 53 points indicating the presence of depression and higher scores indicating severe depression. Patients were also categorised as having comorbid dysphoric moods (anxiety and or depression) and not having comorbid dysphoric moods according to their scores.Additionally, the following data were collected: medical history and new diagnoses of diabetes or impaired glucose tolerance (IGT), dyslipidemia, hyperuricemia, and family history of hypertension. Diabetes was diagnosed based on the presence of typical symptoms (e.g., excessive thirst, polyuria, polyphagia, unexplained weight loss), with confirmation by one of the following criteria: random plasma glucose ≥ 11.1 mmol/L, fasting plasma glucose ≥ 7.0 mmol/L, 2-hour plasma glucose ≥ 11.1 mmol/L during an oral glucose tolerance test (OGTT), or HbA1c ≥ 6.5% [11, 12]. In the absence of typical symptoms, diagnosis required confirmation by repeat testing. Impaired glucose tolerance (IGT) was defined as a 2-hour plasma glucose level between 140Ìýmg/dL (7.8 mmol/L) and 199Ìýmg/dL (11.1 mmol/L) during an OGTT [12]. Dyslipidemia was diagnosed if at least one of the following criteria was met: serum total cholesterol ≥ 5.2 mmol/L (200Ìýmg/dL), triglycerides ≥ 1.7 mmol/L (150Ìýmg/dL), low-density lipoprotein cholesterol ≥ 3.4 mmol/L (130Ìýmg/dL), or high-density lipoprotein cholesterol < 1.0 mmol/L (40Ìýmg/dL) [11]. Hyperuricemia was defined as serum uric acid levels > 420 µmol/L (7.0Ìýmg/dL) in men or > 360 µmol/L (6.0Ìýmg/dL) in women [13]. Family history of hypertension was defined as having a first-degree relative (i.e., parent, sibling, or other direct relative) who had been diagnosed with hypertension.

Observation indicators

The general information from the hypertensive and non-hypertensive groups of the training dataset was compared using univariate analysis. Indicators with statistically significant differences were further analysed using Lasso regression and introduced into the multifactorial logistic regression analysis for analysing the correlation of essential hypertension in young and middle-aged people with screen time spent watching short videos at bedtime, as well as the independent predictors of hypertension in these patients. Based on the logistic multifactorial regression results, we established a nomogram prediction model, and assessed its predictive efficacy by the area under receiver operating characteristic (ROC) curve (AUC). For the plotted ROC curve, the model had a higher predictive efficacy if AUC was closer to 1.0. Curve fit was determined by drawing a calibration plot. Validation datasets were utilized for validating nomogram, ROC and calibration curves were plotted, AUC was calculated, and calibration curve fitness was observed. Clinical application value of our nomogram was analysed by plotting a decision curve.

Statistical analysis

All data were collected in duplicate and recorded in an EXCEL form. R programming language was used for statistical analysis and processing. Measurement data were reported as means ± standard deviations (± s). Comparison between data with a normal distribution and uniform variance was made by t-test. Data that were not normally distributed were expressed as M (P25, P75), and comparison between two groups was accomplished by Mann–Whitney U-test. Count data were analysed and compared via chi-square test. The training set was screened for hypertension risk factors among young and middle-aged adults using Lasso regression, and the statistically significant risk predictors were introduced into the multifactorial logistic model for further analysis. Independent risk factors were screened out, and the ‘rms’ package in R programming language was exploited for creating a nomogram-based risk predictive model for hypertension. Hosmer–Lemeshow test was employed to examine the goodness of fit of the created model. The ‘proc’, ‘car’, and ‘rada’ packages were used to plot the ROC, calibration, and decision curves, with which the validation set was analysed. All statistics were analysed using a two-tailed test, and the significance was evaluated at P &±ô³Ù; 0.05.

Results

Univariate analysis

In the training dataset of 3023 participants, the screen time spent watching short videos at bedtime was as follows: 0Ìýh by 257 (8.50%), 0 < time ≤ 1Ìýh by 633 (20.9%), 1 < time ≤ 2Ìýh by 807 (26.7%), 2 < time ≤ 3Ìýh by 644 (21.3%), 3 < time ≤ 4Ìýh by 412 (13.6%), and > 4Ìýh by 270 (8.93%) participants. Comparison of the general information from the hypertensive and non-hypertensive groups revealed a statistically significant inter-group disparity regarding the screen time spent watching short videos at bedtime (P < 0.05). Additionally, systolic blood pressure (mmHg) measured during health checkups was significantly higher in the hypertensive group (130.51 ± 10.59) compared to the non-hypertensive group (122.42 ± 9.71, P < 0.001), and diastolic blood pressure (mmHg) was also significantly higher in the hypertensive group (86.06 ± 6.37) compared to the non-hypertensive group (76.57 ± 6.60, P < 0.001).Moreover, statistical significance was noted for inter-group disparities (P < 0.05) based on sex, age, occupation, high-sodium diet, physical activity, sleep, overweight or obesity, diabetes or impaired glucose tolerance, dyslipidaemia, hyperuricaemia, and family history. Insignificant inter-group disparities (P > 0.05) were observed when smoking, alcohol consumption, and comorbid dysphoric moods were compared (TableÌý1).

Table 1 Comparison of the general information between hypertensive and non-hypertensive groups

Lasso regression analysis

Statistically significant parameters (P < 0.05) in the univariate assessment were screened using Lasso regression. The results showed lambda_1se = 0.01379897 and lambda_min = 0.001782205. We screened 12 non-zero coefficient correlates for sex, age, screen time, occupation, high-sodium diet, physical activity, sleep, overweight or obesity, diabetes or impaired glucose tolerance, dyslipidaemia, hyperuricaemia, and family history (Fig.Ìý2).

Fig. 2

Screening of Lasso regression variables screening

Multifactorial analysis

Multifactorial logistic regression was performed on the 12 indicators screened using Lasso regression with the following results: screen time of 0 < time ≤ 1Ìýh (odds ratio [OR]: 3.461, 95% CI: 2.022–6.082, P < 0.05), 2 < time ≤ 3Ìýh (OR: 2.645, 95% CI 1.538–4.665, P < 0.05), 3 < time ≤ 4Ìýh (OR: 9.3, 95% CI: 5.327–16.691, P < 0.05), and time > 4Ìýh (OR: 40.236, 95% CI: 21.382–78.15, P < 0.05), suggesting that the screen time spent watching short videos at bedtime is correlated with essential hypertension among young and middle-aged individuals. Other risk factors included being a worker (OR: 1.816, 95% CI: 1.234–2.676); being a professional and technical personnel (OR: 2.639, 95% CI: 1.908–3.668); high-sodium diet (OR: 4.319, 95% CI 3.243–5.804); being overweight or obese (OR: 2.160, 95% CI: 1.68–2.783); having diabetes or impaired glucose tolerance (OR: 2.774, 95% CI: 2.041–3.781); having dyslipidaemia (OR: 1.535,95% CI: 1.196–1.97); having hyperuricaemia (OR: 2.227, 95% CI: 1.686–2.947); age (OR: 1.117, 95% CI: 1.010– 1.134); and family history (OR: 2.665,95% CI 2.094–3.4) (all P < 0.05). Protective factors included being female (OR: 0.264, 95% CI: 0.204–0.34), freelance work (OR: 0.372, 95% CI: 0.193–0.688), unemployment (OR: 0.124:95% CI: 0.058–0.249), and engaging in sufficient levels of physical activity (OR: 0.399, 95% CI: 0.313–0.507) (all P < 0.05) (TableÌý2).

Table 2 Results of multifactor logistics regression analysis

Establishment and evaluation of the nomogram for hypertension risk

On the basis of multifactorial logistic regression results, a nomogram of essential hypertension among local young and middle-aged people was constructed by combining the following influencing factors: sex, age, screen time, occupation, high-sodium diet, adequate physical activity, adequate sleep, overweight or obesity, diabetes mellitus or glucose intolerance abnormalities, dyslipidaemia, hyperuricaemia, and family history (Fig.Ìý3). Each independent predictor was projected upward to the value of the topmost ‘point’ to yield a score of 0–100, and the overall score was documented for accurately forecasting hypertension risk in the corresponding middle-aged youth. A greater overall score indicated a higher probability of developing hypertension. The accuracy, sensitivity, specificity, and AUC of the training dataset for predicting the occurrence of essential hypertension among young and middle-aged populations were 0.864, 0.832, 0.877, and 0.934 (95% CI: 0.925–0.943), respectively (Fig.Ìý4), demonstrating that the nomogram has a preferable discriminability. The calibration plot confirms a good agreement between the forecasted and actual risks (Fig.Ìý5).

Fig. 3

Nomogram of the primary hypertension risk prediction model for young and middle-aged adults

Fig. 4

Receiver operator characteristic curve of the nomogram for predicting localised essential hypertension in young and middle-aged individuals (ACC = 0.864, sensitivity = 0.832, specificity = 0.877)

Fig. 5

Calibration curve of the nomogram model for predicting essential hypertension in local young and middle-aged adults

Validation of the constructed nomogram for predicting the risk of essential hypertension in young and middle-aged adults

Test dataset was used to validate the nomogram model, and the accuracy, sensitivity, specificity, and AUC of the test dataset for predicting the hypertension occurrence among young and middle-aged individuals were 0.864, 0.832, 0.877, and 0.911 (95% CI: 0.895–0.928), respectively (Fig.Ìý6), indicating that the prediction model had good discriminative properties. The calibration curves indicated that the nomogram had good consistency and fit (Fig.Ìý7).

Fig. 6

Test set ROC

Fig. 7

Test set calibration curve

Decision curve analysis

The hypertension predictability of our nomogram for young and middle-aged individuals was demonstrated through the decision curve assessment, with horizontal line indicating no intervention and a zero net benefit, and diagonal line indicating exposure of the entire patients to intervention. The wide spectrum of high-risk threshold probabilities obtained from the decision curve analysis is applicable to both the training and test datasets, suggesting that the nomogram model has clinical applications (Fig.Ìý8).

Fig. 8

Decision curve analysis

Discussion

In the present cross-sectional research, data of young and middle-aged people who underwent medical checkups at Hengshui People’s Hospital were utilized for analysing the association between essential hypertension and the time spent watching short videos at bedtime in young and middle-aged adults in Hengshui City, China, and found that the time spent watching short videos at bedtime was linked significantly to essential hypertension among the study population in Hengshui City at levels of 0 < time ≤ 1Ìýh, 2 < time ≤ 3Ìýh, 3 < time ≤ 4Ìýh, and time > 4Ìýh. Among them, the screen time OR of > 4Ìýh reached 40.236%, which was possibly related to the fact that the study population was a medical checkup population and the inclusion criteria were stringent, resulting in selective bias. However, time spent watching short videos at bedtime of 1 < time ≤ 2Ìýh was not linked to essential hypertension among the investigated population. Overall, these findings suggest that for young and middle-aged individuals, longer screen time spent watching short videos at bedtime may be associated with a higher hypertension prevalence. We combined the time spent watching short videos at bedtime with other common risk factors for hypertension to construct a nomogram that was internally validated. The nomogram demonstrated good discrimination and calibration. Our nomogram exhibited a C-index of 0.934, suggesting that the model has good application value for predicting essential hypertension among young and middle-aged people.

Our results agree with those of a cross-sectional research from Sweden, which subjected the cross-sectional information of over 45,000 males and females from two Swedish cohort investigations, LifeGene (18–45 years) and EpiHealth (45–75 years), to linear regression assessment. As shown by the results, prolonged television screen time was linked to higher systolic and diastolic BP levels [6]. Another prospective study in a paediatric cohort from Odense demonstrated a significant cross-sectional correlation of screen time at bedtime from preschool onward with hypertension [14]. To explain the correlation of screen time with hypertension, several mechanisms have been put forward, mainly focusing on the idea that screen time is sedentary behaviour, with a meta-analysis encompassing 28 studies concluding that an extra hour of sitting per day is linked to 0.06- and 0.2-mmHg elevations in systolic and diastolic BP levels, respectively, as well as a 1.02 heightening in the hypertension odds ratio [15]. While traditional screen time encompasses the time spent watching television, playing video games and using computers, for example, people may watch television accompanied by a certain amount of physical activity, our study was based on screen time spent watching short videos at bedtime, which is more reflective of a sedentary nature. Moreover, watching short videos before bedtime can cause sympathetic arousal [16], which may be another important mechanism for the BP elevation caused by watching short videos at bedtime.

Our findings are inconsistent with those of a U.S. longitudinal prospective cohort research, which showed that every extra hour of screen time per day was linked to an internal increase in body mass index; and to the internal odds of obesity, large waist circumference and diabetes mellitus; and that screen time did not correlate significantly with hypertension or hyperlipidaemia [5]. The reasons for analysis may be related to this study’s insufficiency of diversity in screen time measurements and the fact that some metabolic traits, including BP, cholesterol and haemoglobin A1c, were measured only in later waves (waves 4 and 5), which may have limited the analysis on association of the screen time with the risk of cardiometabolic disorders over time.

We also analysed other common risk factors for essential hypertension and constructed a nomogram-based prediction model. The prevalence of hypertension tends to increase with age [17], and sex-based differences have been observed in the prevalence characteristics of the disease. Hypertension is less prevalent among young women than among men. Moreover, hypertension is more prevalent among females than in males of identical age after 60 years of age [18]. Hypertension is significantly correlated with occupational stress [19], and salt intake is dose-dependently associated with the elevated BP [8]. For Patients with hypertrnsion, sports activities and exercise can lower their BP [20]. Persistent sleep duration of < 7–8Ìýh is linked to the hypertension onset [10]. Loss of weight in adults reduces their risk of developing hypertension [21, 22]. Hypertension correlated with diabetes mellitus, which is linked closely to a heightened incidence of cardiovascular complications [23]. Dyslipidaemia is an independent predictor of hypertension, and patient lipid molecules are associated with increased cardiovascular risk by affecting the elasticity and structure of arterial vasculature, causing elevated blood pressure [24, 25]. Blood uric acid can elevate blood pressure by inducing vascular endothelial dysfunction, activating renin-angiotensin-aldosterone, stimulating vascular smooth muscle proliferation, and decreasing nitric oxide production [26, 27]. The probability of developing hypertension is elevated by presence of a family history [28]. This study suggests that sex, age, occupation, high-sodium diet, overweight or obesity, diabetes or impaired glucose tolerance, dyslipidaemia, hyperuricaemia, and family history are independently correlated with essential hypertension among young and middle-aged adults and that being female, engaging in sufficient physical activity, and having adequate sleep are protective factors against essential hypertension for the study population, which is in line with previous reports. The multifactorial analysis in this study did not find any correlation between parameters like smoking, alcohol consumption, marital status, literacy attainment and hypertension among young and middle-aged individuals, which was considered to be related to the comparatively small sample size ascribed to the stringent inclusion and exclusion criteria of this study.

Limitations

This research had certain shortcomings. First of all, it is a single-centre investigation, and the study population was local participants who participated in health checkups; therefore, selective bias could not be avoided. Second, the inclusion of risk factors for hypertension was not comprehensive. For example, the study period included the period of the novel coronavirus pandemic, which may have affected the population physiologically, psychologically, socially, and in many other ways [29]. However, this research did not include the new coronavirus epidemic as a risk factor. Third, time spent watching short videos at bedtime was not compared with other periods or other categories of screen time. Fourth, screen time in this study was self-reported and not assessed for reliability, which may have introduced recall and reporting bias. Fifth, only total daily sleep time was assessed when sleep was evaluated, and sleep quality was not assessed. Sixth, during the survey, we observed that participants were either unable to accurately recall or were unwilling to report the specific content of the short videos they had watched. The varying content of these videos could potentially have differential effects on participants’ blood pressure. Seventh, the nomogram-based prediction model was only internally validated and lacked external validation results from other centres. Therefore, subsequent clinical data from larger samples and multiple centres are needed to further clarify the correlation between the time spent watching short videos at bedtime and essential hypertension among young and middle-aged individuals and actively explore factors influencing essential hypertension for these age groups to optimise the nomogram-based prediction model.

Conclusions

For young and middle-aged adults, the time spent watching short videos at bedtime, as well as sex, age, occupation, high-sodium diet, overweight or obesity, diabetes or impaired glucose tolerance, dyslipidaemia, hyperuricaemia, and family history are associated with essential hypertension, and our nomogram-based prediction model of hypertension risk established by multifactorial logistic regression has good predictive efficacy. This suggests that strict control of screen time spent watching short videos at bedtime, control of body weight, blood lipids, blood glucose, and uric acid levels, and improvement of poor lifestyle, such as a high-sodium diet, are conducive to the reduction of the hypertension occurrence and development among young and middle-aged individuals. Besides, our nomogram can be used to screen and implement early intervention for high-risk groups, thus lowering the morbidity rate and enhancing the long-term prognosis for this population.