Online health information seeking,healthcare utilization,and exercise-related self-management among patients with long-term conditions in China during COVID-19

Abstract

Objective

This study explored the associations between online health information seeking (OHIS), healthcare utilization, and exercise-related self-management behaviors among adults in China during the COVID-19 pandemic, focusing on individuals with long-term conditions (LTCs). It was guided by the biopsychosocial model and the Information-Motivation-Behavioral Skills (IMB) model.

Methods

A cross-sectional analysis used observational data from 1,831 respondents in the 2021 China General Social Survey (CGSS). OHIS was defined as the frequency of using the internet to obtain health or medical information in the past 12 months. Healthcare utilization was measured by the frequency of medical visits, including both traditional Chinese and Western medicine. Exercise-related self-management was represented by regular physical exercise. Multinomial logistic regression was applied while controlling for demographic, psychosocial, and health-related factors.

Results

OHIS adopters reported on average more frequent medical visits and higher levels of physical exercise than non-adopters. Meanwhile, among individuals with LTCs, OHIS is associated with less frequent medical visits but a higher likelihood of physical exercise, suggesting a potential pathway linking OHIS to exercise-related self-management behavior.

Conclusion

OHIS was positively associated with healthcare utilization and exercise-related self-management behaviors during a period of restricted healthcare access in China. These findings suggest that accessible and reliable online health information may complement patients’ exercise-related self-management capacities in developing countries, offering insights for integrating digital health strategies into primary care and LTCs management.

Keywords

online health information seeking long-term conditions healthcare service utilization exercise-related self-management COVID-19 multinomial logit model

Introduction

Globally, Non-communicable diseases (NCDs) account for 75% of all deaths.¹ Long-term conditions (LTCs) impose an even broader burden, encompassing NCDs as well as functional impairments, mental health disorders, and post-infection syndromes, placing substantial strain on individuals and healthcare systems.² In England, around 15.4 million people lived with LTCs in 2018, with prevalence among those aged 60 and above projected to reach 58% by 2035.³ In developed countries such as the United States, promoting patient self-management has emerged as a core strategy for LTCs management.^4,5 Effective self-management enhances medication adherence, supports healthier lifestyles and allows early detection of symptom worsening. Ultimately, these benefits lead to better quality of life, improved health outcomes, and lower healthcare costs, which help ease pressure on the healthcare system.^6,7

In developing countries such as China, the self-management of LTCs remains a significant challenge. The prevalence of LTCs in China has increased markedly, rising from 17.0% in 1993 to 34.3% in 2018.⁸ Meanwhile, limited primary healthcare capacity further constrains self-management support. Although over 40% of residents have signed contracts with family doctors, these services often focus on basic follow-ups and lack personalized management, thereby reducing their effectiveness in guiding patients with LTCs.^9,10 Basically, patients lack the necessary skills and support for effective self-management.

Facing these systemic constraints, patients increasingly turn to alternative information channels.^11,12 Studies show that around 80% of patients with LTCs have sought health information online, with this group exhibiting significantly higher OHIS frequency than the general population.¹³ High-quality online information supports health maintenance, improvement, and restoration^14–16 and can reduce perceived risk, alleviate health-related anxiety, and provide emotional support, for example in cancer-related searches.¹⁷ Therefore, for many developing countries, OHIS may serve as a cost-effective strategy to address health information gaps in self-management.¹²

However, OHIS also comes with potential risks. The uneven quality of online health information may lead users to make misguided health decisions.¹⁸ Individuals with lower eHealth literacy are particularly vulnerable to information overload, which can trigger health anxiety or cyberchondria and drive unnecessary healthcare utilization, thereby wasting resources and widening health inequalities.^19–22 Moreover, not all patients with LTCs actively seek online health information.⁴ According to the biopsychosocial model, their heightened health anxiety may foster information avoidance, resulting in delayed treatment, poor self-management, and aggravated psychological distress.^23–26

Existing research shows a significant positive association between OHIS and healthcare service utilization, with notable heterogeneity across subgroups.^27,28 For instance, female OHIS adopters are 1.42 times more likely to seek medical services than non-adopters.²⁸ OHIS also markedly improves medical resource access for those aged 70 and older compared to the 50–69 age group.²⁹ And some highlights positive effects of OHIS, such as reduced information asymmetry and better-informed medical decisions.^27,28,30 Others point to negative outcomes like information overload or overestimation of care benefits.³¹ Nevertheless, these mechanisms remain theoretically and empirically underexplored.

The COVID-19 pandemic has provided an ideal research situation for exploring this mechanism. Based on data from the 2021 China General Social Survey, conducted from June to September 2021 during a critical period of China’s dynamic zero-COVID policy (including strict measures such as inter-provincial travel restrictions and isolation for close contacts).³² During this unique period, (1) uncertainty brought by the novel virus, social isolation, psychological stress, and the “Infodemic” had expanded health anxiety beyond specific risk groups to become a widespread psychological phenomenon^33–35; (2) the implementation of lockdown policies in countries like China, the UK, and Italy had resulted in restricted access to routine medical services for non-COVID patients^36,37 and the implementation of policies like “avoid non-essential medical visits”^38,39 had restructured the healthcare system. The combination of these two factors allows for a more isolated assessment of OHIS effects while minimizing non-essential healthcare-seeking conditions.

Furthermore, the absence of the usual support from the healthcare system has underscored the vital importance of self-management for patients with LTCs.^40–42 However, not all patients possess the necessary health information to effectively manage their conditions.⁴ In this context, online health information emerged as an indispensable resource due to its accessibility, convenience, and contact-free nature during periods of constrained healthcare access.^4,13,17

Guided by the biopsychosocial model⁴³ and the Information-Motivation-Behavioral Skills (IMB) model,⁴⁴ this study conceptualizes the mechanisms linking OHIS, healthcare service utilization, and exercise-related self-management among patients with LTCs. Within the IMB framework, OHIS serves as a critical source of information that shapes health behaviours through two distinct pathways. First, reliable online information enhances motivation and self-efficacy, thereby equipping patients with the behavioural skills necessary for effective exercise-related self-management and rational healthcare utilization.^4,45 Conversely, exposure to low-quality or overwhelming information may distort risk perceptions and induce health anxiety, leading to maladaptive behaviours such as unnecessary healthcare-seeking or avoidance.^46,47

Furthermore, consistent with the Biopsychosocial model,⁴³ biological context (specifically Long-term condition status) moderates these relationships.^4,48 We posit that for patients with LTCs, the association between OHIS and medical visit frequency is attenuated, as their established care routines reduce the need for new information to drive utilization. In contrast, the relationship between OHIS and exercise-related self-management is likely stronger in this group, as the emotional and informational support derived from online sources reinforces the motivation required for sustained disease management.^4,49

This study aims to advance understanding of digital health’s role in supporting exercise-related self-management among patients with LTCs in developing countries. Although digital health initiatives have expanded globally, empirical evidence from these settings remains limited. By examining OHIS as a low-cost, accessible approach to bridge healthcare information gaps, this research highlights a practical strategy to promotes digital health equity and hence to enhance patient empowerment. The findings provide empirical reference and actionable insights for policymakers and practitioners on leveraging digital tools to strengthen exercise-related self-management capacity, improve care quality, and promote health equity in developing economies.

Methodology

Data

This study used secondary data from the 2021 China General Social Survey (CGSS), a nationally representative cross-sectional observational survey conducted by the National Survey Research Center at Renmin University of China (NSRC). The 2021 survey was implemented across most provinces in mainland China between July and December 2021 to collect information from adults aged 18 and above. All data were collected by trained interviewers through face-to-face interviews. The study protocol of the CGSS was reviewed and approved by the Institutional Review Board of the NSRC, and the data collection followed ethical standards with written informed consent obtained from all participants, and data were fully anonymized before being released for public use.

The 2021 survey included core modules, thematic modules, and the International Social Survey Programme (ISSP) Health Module. The core and thematic modules were administered to all respondents (N = 8,148), and the ISSP Health Module was completed by a random one-third subsample (N = 2,690). All participants in this study were respondents from the ISSP Health Module.

The ISSP Health Module contains standard items developed collaboratively within the ISSP framework, assessing constructs such as self-rated health through single or multi-item Likert-type scales. These measures have demonstrated good cross-national reliability and construct validity in previous research.^50,51 After excluding incomplete or invalid responses (refusals to answer or indicating “don’t know”), the final analytic sample consisted of 1,831 respondents. Given the cross-sectional and observational design of the CGSS, the study findings describe associations.

Outcome variables

Building on the biopsychosocial and IMB models described in the introduction, we operationalized the conceptual framework into an empirical analytical model linking OHIS to health-related behavioral outcomes. In the IMB framework, OHIS represents the information component that can strengthen or undermine motivation and behavioral skills. These mechanisms are reflected empirically through two behavioral outcomes that together capture the behavioral skill dimension: (1) healthcare service utilization, measured by the frequency of medical visits, and (2) exercise-related self-management behaviors, represented by the frequency of physical exercise. These relationships vary depending on the presence of LTCs, consistent with the biopsychosocial perspective that LTCs shape patterns of health information use and behavioral response. A range of demographic, health status, and health resource variables were included as controls to account for individual differences in the biopsychosocial context.

Dependent variable

The first dependent variable was an ordered measure of the medical visits’ frequency in the last year. Given that the China’s healthcare system treats traditional Chinese medicine (TCM) and Western medicine as equally important,⁵² and to simplify the analytical model, the variable combined two survey questions regarding respondents’ frequency of visits to TCM practitioners and Western medical doctors over the past year. We computed the average of the responses to these two questions and categorized them into four groups: “never,” “seldom,” “sometimes,” and “often.”

The second dependent variable measures the frequency of physical exercise. The variable was constructed from a question that asked whether a participant regularly participated in physical exercise during their leisure time.⁵³ The response to the question was “never,” “a few times a year,” “several times a month,” “several times a week,” or “daily.”

The frequency of medical visits measures healthcare service utilization. It is known that healthcare service utilization refers to the frequency and intensity of an individual’s use of medical services during a specific period.⁵⁴ As a fundamental measure of healthcare system effectiveness, healthcare service utilization frequently serves as a key dependent variable in health behavior research.⁵⁵

The variable of the frequency of physical exercise serves as an indicator of one aspect of health management behavior. In the context of patients with LTCs, self-management emphasizes the central role of patients in managing their health, encompassing “medical or behavioral management of the disease, role management, and emotional management.”⁵⁶ This underscores the physical exercise as a fundamental component of self-management.⁶ Additionally, for individuals without LTCs, physical exercise is also crucial for disease prevention.

Independent variables

The key variable of interest, health information seeking, was constructed from the question “During the past 12 months, how often, if at all, did you use the internet on any device (such as computers, tablets and smartphones) to look for health or medical information for yourself or someone else?”⁵⁷ The responses to the questions were “never used,” “rarely used,” “several times a day,” “once a day,” “several times a week,” “several times a month” and “several times a year.” We subsequently dichotomized the responses into “non-adopter” and “adopter,” where “non-adopter” includes “never used” and “rarely used.”

Another independent variable identifies the status of LTC, a dichotomous measure of whether a participant reported as a patient with LTCs. The variable was constructed from questions regarding whether a participant had any long-standing illnesses, chronic diseases, or disabilities. Those answering affirmatively were classified as patients with LTCs.⁵⁸

Control variables

(1) Demographics: Age (≥65=1), gender (male=1), education (primary or below=0, middle school=1, undergraduate or above =2), income (zero income=0, low income=1, medium income=2, high income=3), and marital status (married=1). These demographic characteristics which are known to influence OHIS,⁵⁹ were controlled for in this study. And living situation (living alone=0, living with family=1) as an important influencing factor during the lockdown period of the COVID-19, have also been controlled.

(2) Health status & behaviors: health anxiety (“I am very distressed about my health condition.” No=0), unhealthy behavior (“Do you currently smoke? Do you drink four or more units of alcohol per day?” Both no=0, smoking or alcohol=1, both smoking and alcohol=2), and self-rated health (very poor=0, poor=1, average=2, good=3, excellent=4).

(3) Health beliefs: vaccination belief (responses agreeing with “Overall, the disadvantages of vaccination outweigh the advantages” were recoded as negative vaccination belief = 0), trust in doctors (trust=1), and perceived utility of online health information (responses agreeing with “The Internet can help people determine whether they need to see a doctor.” were recoded as positive=1).

(4) Health insurance status: uninsured=0, basic health insurance=1, commercial/supplementary/other health insurance=2, hold two or more types=3.

(5) Information technology environment: internet usage (used = 1) and living in a smart-plan city (yes = 1). A more detailed description is provided in Supplemental Table 1.

Statistical analysis

Multinomial logistic regression was employed to examine two primary outcomes, both the frequency of medical visits and the frequency of physical exercise. Both outcomes comprise ordered categories (e.g., never, seldom, sometimes, often). The proportional odds assumption was tested using the Brant test and was violated for both outcomes based on the observed data structure. Therefore, multinomial logistic regression was chosen as it accommodates polytomous outcomes without the proportional odds constraint, enabling category-specific variation in relative risk ratios.^60,61 This flexibility provides a more refined understanding of how predictors (e.g., OHIS and LTCs) differentially associate with distinct levels of the outcomes, beyond what ordinal models can capture.^62,63

Analyses proceeded in two stages. First, main-effects models estimated associations of OHIS adoption and LTCs with each outcome, controlling for control variables. Second, interaction terms between OHIS and LTCs were introduced to test. Weights and design effects were not applied in this study. All analyses were conducted in Stata 18.

Results

Descriptive statistics

As reported in Table 1, the sample characteristics of 1,831 survey respondents closely aligned with the national population structure, confirming the data’s national representativeness. Specifically, 18.8% of respondents were aged 65 and above, closely mirroring the 2020 national statistic of 17.59%, health insurance coverage rate in the sample was 95.7%, consistent with the national rate of 96%. Approximately 76.7% of respondents reported at least a middle school education, including 25.0% with an undergraduate degree or higher, which slightly exceeded the national level of 20.7%.

Table 1.

Descriptive statistics classified by OHIS adoption status.

Variable	Full sample (N, %)	Non-adopter	Adopter	P-value
	1831	800 (43.7%)	1031 (56.3%)
Frequency of medical visits
Never	525 (28.7%)	268 (33.5%)	257 (24.9%)	<.001
Seldom	779 (42.5%)	298 (37.2%)	481 (46.7%)
Sometimes	420 (22.9%)	168 (21.1%)	252 (24.4%)
Often	107 (5.9%)	66 (8.2%)	41 (4.0%)
Frequency of physical exercise
Never	538 (29.4%)	360 (45.0%)	178 (17.3%)	<.001
A few times a year	231 (12.6%)	77 (9.6%)	154 (14.9%)
Several times a month	266 (14.5%)	59 (7.4%)	207 (20.1%)
Several times a week	327 (17.9%)	101 (12.6%)	226 (21.9%)
Daily	469 (25.6%)	203 (25.4%)	266 (25.8%)
Long-term conditions
Without	1292 (70.6%)	478 (59.8%)	814 (79.0%)	<.001
With	539 (29.4%)	322 (40.2%)	217 (21.0%)
Demographics
Gender
Female	996 (54.4%)	425 (53.1%)	571 (55.4%)	.34
Male	835 (45.6%)	375 (46.9%)	460 (44.6%)
Age
<65	1486 (81.2%)	547 (68.4%)	939 (91.1%)	<.001
≥65	345 (18.8%)	253 (31.6%)	92 (8.9%)
Education
Primary and below	426 (23.3%)	311 (38.9%)	115 (11.2%)	<.001
Middle school	947 (51.7%)	407 (50.9%)	540 (52.4%)
Undergraduate and above	458 (25.0%)	82 (10.2%)	376 (36.4%)
Income
Zero income	342 (18.7%)	157 (19.6%)	185 (18.0%)	<.001
Low income	517 (28.3%)	314 (39.3%)	203 (19.7%)
Medium income	486 (26.5%)	199 (24.9%)	287 (27.8%)
High income	486 (26.5%)	130 (16.2%)	356 (34.5%)
Marital status
Unmarried	445 (24.3%)	179 (22.4%)	266 (25.8%)	.09
Married	1386 (75.7%)	621 (77.6%)	765 (74.2%)
Living situation
Living alone	464 (25.3%)	209 (26.1%)	254 (24.6%)	.47
Living with family	1368 (74.7%)	591 (73.9%)	777 (75.4%)
Health status & behaviors
Health anxiety
No anxiety	1260 (68.8%)	514 (64.2%)	746 (72.4%)	<.001
Anxiety	571 (31.2%)	286 (35.8%)	285 (27.6%)
Unhealthy behavior
None	1329 (72.6%)	572 (71.4%)	758 (73.5%)	.36
Smoking or alcohol	346 (18.9%)	163 (20.4%)	184 (17.8%)
Both smoking and alcohol	156 (8.5%)	66 (8.2%)	90 (8.7%)
Self-rated health
Very poor	63 (3.4%)	44 (5.5%)	19 (1.9%)	<.001
Poor	191 (10.4%)	132 (16.5%)	59 (5.7%)
Average	544 (29.7%)	249 (31.1%)	295 (28.6%)
Good	677 (37.0%)	249 (31.1%)	428 (41.5%)
Excellent	356 (19.4%)	126 (15.8%)	230 (22.3%)
Health belief
Vaccination belief
Negative belief	311 (17.0%)	148 (18.5%)	163 (15.8%)	.13
Positive belief	1522 (83.0%)	652 (81.5%)	868 (84.2%)
Trust in doctors
No trust	195 (10.6%)	93 (11.6%)	102 (9.9%)	.24
Trust	1636 (89.4%)	707 (88.4%)	929 (90.1%)
Perceived utility of online health information
Negative	827 (45.2%)	413 (51.6%)	414 (40.2%)	<.001
Positive	1004 (54.8%)	387 (48.4%)	617 (59.8%)
Health insurance status
Uninsured	78 (4.3%)	48 (6.0%)	30 (2.9%)	<.001
Basic health insurance	1395 (76.2%)	642 (80.2%)	753 (73.1%)
Commercial/supplementary/other	93 (5.1%)	36 (4.5%)	57 (5.5%)
Hold two or more types	265 (14.5%)	74 (9.2%)	191 (18.5%)
Information technology environment
Internet usage
Not used	322 (17.6%)	305 (38.1%)	17 (1.6%)	<.001
Used	1509 (82.4%)	495 (61.9%)	1014 (98.4%)
Smart city
No	860 (47.0%)	394 (49.3%)	466 (45.2%)	.09
Yes	971 (53.0%)	406 (50.7%)	565 (54.8%)

Further statistical analysis indicated that 56.3% (1031) of respondents had adopted online health information seeking, while 43.7% (800) had not. 42.5% of respondents had relatively low annual medical visits frequencies, with only 5.9% requiring frequent medical visits. The proportion of non-adopter (OHIS) group with “never” exercising was the highest (45.0%, n=360), while the distribution of exercise frequencies within the OHIS group was relatively balanced (15%-25%). The proportion of “daily” exercisers was very similar between the two groups (non-adopter 25.4% vs. adopter 25.8%) and aligned with the overall distribution trend. A total of 539 (29.4%) respondents indicated having LTCs, with 217 (21.0%) of the adopters reporting LTCs. Likelihood ratio chi-square test results indicated significant differences between the non-adopters and the adopters in terms of frequency of medical visits, frequency of physical exercise, long-term conditions, health anxiety, and health insurance status.

Patients with LTCs exhibit higher healthcare utilization rates, with 35.6% reporting “sometimes” seeking medical care and 14.7% reporting “often,” compared to only 17.6% and 2.2% among individuals without LTCs, respectively. This highlights the high frequency of medical visits among patients with LTCs. Additionally, patients with LTCs engage in less physical exercise, as 37.7% of them never exercise (compared to 25.9% of individuals without LTCs), and only 12.8% exercise “several times a week” (compared to 20.0%). However, the proportion of patients with LTCs who exercise “daily” (28.8%) is slightly higher than that of individuals without LTCs (24.3%). (see Supplement Table 2 for details).

Analysis of the impact of OHIS on the frequency of medical visits and physical exercise

The Mlogit results reveal that OHIS significantly increases the likelihood of more frequent medical visits (Table 2). Compared with non-adopters, OHIS adopters were nearly twice as likely to make seldom visits (RRRs = 1.949, p < .001) and almost three times as likely to make Sometimes visits (RRRs = 2.822, p < .001), though the association was not significant for Often visits (p = .44). Similarly, individuals with LTCs showed a markedly higher probability of seeking medical care, being over twice as likely to make Sometimes visits (RRRs = 2.264, p < .001) and more than four times as likely to make Often visits (RRRs = 4.384, p < .001) than respondents without LTCs. Other sociodemographic variables, including age, income, and health anxiety also showed partial significance, while education, gender, and psychosocial factors (e.g., trust in doctors) exhibited no consistent associations. And better self-rated health showed a negative association.

Table 2.

Analysis of factors influencing the frequency of medical visits (multinomial logistic regression).

	Seldom			Sometimes			Often
	RRRs (Standard error)	P-values	95% CI	RRRs (Standard error)	P-values	95% CI	RRRs (Standard error)	P-values	95% CI
OHIS
Non-adopter	Reference			Reference			Reference
Adopter	1.949 (0.266)	<.001	(1.492, 2.545)	2.822 (0.499)	<.001	(1.995, 3.993)	1.254 (0.368)	.44	(0.706, 2.228)
LTCs
Without	Reference			Reference			Reference
With	1.381 (0.243)	.07	(0.978, 1.950)	2.264 (0.440)	<.001	(1.547, 3.312)	4.384 (1.337)	<.001	(2.412, 7.970)
Gender
Female	Reference			Reference			Reference
Male	0.981 (0.147)	.90	(0.732, 1.316)	0.824 (0.148)	.28	(0.579, 1.171)	0.910 (0.258)	.74	(0.522, 1.587)
Age
<65	Reference			Reference			Reference
≥65	1.073 (0.208)	.71	(0.735, 1.568)	1.981 (0.426)	<.001	(1.300, 3.020)	1.230 (0.395)	.52	(0.656, 2.307)
Education
Primary and below	Reference			Reference			Reference
Middle school	0.740 (0.123)	.07	(0.535, 1.023)	0.807 (0.159)	.28	(0.548, 1.188)	0.867 (0.261)	.64	(0.481, 1.564)
Undergraduate and above	1.048 (0.223)	.83	(0.690, 1.590)	1.107 (0.288)	.69	(0.665, 1.844)	1.778 (0.802)	.20	(0.735, 4.305)
Income
Zero income	Reference			Reference			Reference
Low income	1.111 (0.197)	.55	(0.785, 1.572)	1.362 (0.297)	.16	(0.888, 2.088)	0.958 (0.315)	.90	(0.503, 1.824)
Medium income	1.462 (0.265)	.004	(1.025, 2.085)	2.066 (0.463)	<.001	(1.331, 3.207)	1.821 (0.647)	.09	(0.908, 3.653)
High income	1.936 (0.377)	<.001	(1.322, 2.837)	2.417 (0.597)	<.001	(1.489, 3.920)	1.900 (0.827)	.14	(0.809, 4.458)
Marital status
Unmarried	Reference			Reference			Reference
Married	0.861 (0.131)	.33	(0.638, 1.161)	0.960 (0.181)	.83	(0.663, 1.389)	0.837 (0.248)	.55	(0.468, 1.498)
Living situation
Living alone	Reference			Reference			Reference
Living with family	1.070 (0.154)	.64	(0.808, 1.419)	1.204 (0.215)	.30	(0.849, 1.708)	1.233 (0.360)	.47	(0.696, 2.184)
Health anxiety
No anxiety	Reference			Reference			Reference
Anxiety	1.246 (0.193)	.15	(0.920, 1.688)	2.548 (0.435)	<.001	(1.823, 3.562)	3.442 (0.939)	<.001	(2.016, 5.876)
Unhealthy behavior
None	Reference			Reference			Reference
Smoking or alcohol	0.679 (0.116)	.02	(0.486, 0.948)	0.452 (0.100)	<.001	(0.292, 0.697)	0.554 (0.193)	.09	(0.280, 1.098)
Both smoking and alcohol	0.722 (0.163)	.15	(0.464, 1.123)	0.585 (0.173)	.07	(0.327, 1.044)	0.282 (0.186)	.06	(0.078, 1.026)
Self-rated health
Very poor	Reference			Reference			Reference
Poor	1.735 (0.865)	.27	(0.653, 4.607)	1.660 (0.806)	.30	(0.641, 4.300)	0.670 (0.342)	.43	(0.247, 1.819)
Average	1.342 (0.619)	.52	(0.543, 3.316)	0.872 (0.395)	.76	(0.359, 2.119)	0.296 (0.145)	.01	(0.113, 0.775)
Good	1.062 (0.494)	.90	(0.426, 2.644)	0.563 (0.260)	.21	(0.227, 1.392)	0.116 (0.063)	<.001	(0.040, 0.338)
Excellent	0.696 (0.331)	.45	(0.274, 1.771)	0.350 (0. 170)	.03	(0.135, 0.908)	0.037 (0.029)	<.001	(0.008, 0.169)
Vaccination belief
Negative belief	Reference			Reference			Reference
Positive belief	0.908 (0.148)	.55	(0.659, 1.251)	0.853 (0.166)	.41	(0.582, 1.250)	0.665 (0.195)	.16	(0.375, 1.180)
Trust in doctors
No trust	Reference			Reference			Reference
Trust	1.172 (0.236)	.43	(0.790, 1.738)	0.876 (0.201)	.56	(0.559, 1.373)	0.992 (0.358)	.98	(0.489, 2.012)
Perceived utility of online health information
Negative	Reference			Reference			Reference
Positive	0.822 (0.100)	.11	(0.648, 1.043)	0.760 (0.112)	.06	(0.569, 1.015)	0.950 (0.229)	.83	(0.593, 1.522)
Health insurance status
Uninsured	Reference			Reference			Reference
Basic health insurance	1.341 (0.371)	.29	(0.779, 2.306)	1.680 (0.629)	.17	(0.807, 3.498)	2.888 (1.961)	.12	(0.764, 10.923)
Commercial/supplementary/other	1.544 (0.581)	.25	(0.739, 3.226)	1.426 (0.708)	.48	(0.539, 3.771)	3.011 (2.571)	.20	(0.565, 16.049)
Hold two or more types	0.914 (0.287)	.78	(0.494, 1.690)	1.493 (0.618)	.33	(0.664, 3.362)	2.777 (2.100)	.18	(0.631, 12.225)
Internet usage
Not used	Reference			Reference			Reference
Used	0.787 (0.161)	.24	(0.528, 1.175)	0.763 (0.185)	.27	(0.475, 1.227)	0.803 (0.279)	.53	(0.407, 1.586)
Living in a smart-plan city
No	Reference			Reference			Reference
Yes	1.168 (0.138)	.19	(0.926, 1.473)	1.178 (0.170)	.26	(0.888, 1.564)	1.186 (0.281)	.47	(0.745, 1.887)
Constant	0.783 (0.514)	.71	(0.216, 2.836)	0.145 (0.107)	.01	(0.034, 0.619)	0.106 (0.117)	0.04	(0.012, 0.919)
Observation	1831
R-squared	0.1091

Regarding physical exercise (Table 3), OHIS adoption again demonstrated a strong positive relationship with the frequency of participation. Compared with non-adopters, OHIS users were significantly more likely to engage in exercise across nearly all activity levels, from several times a month (RRRs=3.389, p<.001) to daily participation (RRRs=1.815, p<.001).

Table 3.

Analysis of factors influencing the frequency of physical exercise (multinomial logistic regression).

	A few times a year			Several times a month			Several times a week			Daily
	RRRs (Standard error)	P-values	95% CI	RRRs (Standard error)	P-values	95% CI	RRRs (Standard error)	P-values	95% CI	RRRs (Standard error)	P-values	95% CI
OHIS
Non-adopter	Reference			Reference			Reference			Reference
Adopter	2.578 (0.505)	<.001	(1.756, 3.783)	3.389 (0.687)	<.001	(2.278, 5.042)	2.327 (0.419)	<.001	(1.635, 3.311)	1.815 (0.285)	<.001	(1.334, 2.468)
LTCs
Without	Reference			Reference			Reference			Reference
With	1.126 (0.262)	.61	(0.714, 1.776)	1.101 (0.256)	.68	(0.698, 1.738)	0.920 (0.197)	.70	(0.605, 1.398)	1.338 (0.241)	.11	(0.940, 1.905)
Gender
Female	Reference			Reference			Reference			Reference
Male	1.353 (0.282)	.15	(0.900, 2.036)	1.207 (0.255)	.37	(0.799, 1.826)	1.234 (0.238)	.27	(0.846, 1.800)	0.910 (0.154)	.57	(0.653, 1.267)
Age
<65	Reference			Reference			Reference			Reference
≥65	0.706 (0.193)	.20	(0.413, 1.206)	0.365 (0.124)	<.001	(0.188, 0.710)	0.885 (0.216)	.62	(0.549, 1.426)	1.901 (0.352)	<.001	(1.322, 2.734)
Education
Primary and below	Reference			Reference			Reference			Reference
Middle school	1.208 (0.265)	.39	(0.786, 1.858)	1.597 (0.397)	.06	(0.981, 2.598)	1.742 (0.377)	.01	(1.139, 2.663)	2.306 (0.403)	<.001	(1.637, 3.247)
Undergraduate and above	3.848 (1.245)	<.001	(2.041, 7.255)	8.469 (2.815)	<.001	(4.414, 16.248)	10.834 (3.335)	<.001	(5.926, 19.806)	7.261 (2.078)	<.001	(4.143, 12.724)
Income
Zero income	Reference			Reference			Reference			Reference
Low income	0.921 (0.235)	.75	(0.558, 1.519)	0.743 (0.194)	.26	(0.445, 1.240)	0.762 (0.175)	.23	(0.486, 1.194)	0.999 (0.197)	1.00	(0.679, 1.470)
Medium income	1.157 (0.297)	.57	(0.699, 1.912)	0.765 (0.202)	.31	(0.456, 1.283)	0.797 (0.186)	.33	(0.504, 1.260)	1.131 (0.232)	.55	(0.756, 1.691)
High income	1.214 (0.352)	.50	(0.687, 2.144)	1.562 (0.433)	.11	(0.907, 2.688)	1.034 (0.268)	.90	(0.621, 1.720)	1.127 (0.271)	.62	(0.704, 1.806)
Marital status
Unmarried	Reference			Reference			Reference			Reference
Married	0.519 (0.113)	<.001	(0.339, 0.794)	0.508 (0.110)	<.001	(0.332, 0.777)	0.570 (0.117)	.01	(0.381, 0.852)	0.691 (0.127)	.04	(0.482, 0.990)
Living situation
Living alone	Reference			Reference			Reference			Reference
Living with family	1.017 (0.208)	.93	(0.682, 1.517)	1.283 (0.267)	.23	(0.854, 1.929)	1.297 (0.250)	.18	(0.888, 1.893)	1.062 (0.177)	.72	(0.766, 1.473)
Health anxiety
No anxiety	Reference			Reference			Reference			Reference
Anxiety	1.279 (0.261)	.23	(0.857, 1.909)	1.532 (0.307)	.03	(1.034, 2.269)	1.195 (0.225)	.34	(0.826, 1.729)	1.017 (0.167)	.92	(0.737, 1.403)
Unhealthy behavior
None	Reference			Reference			Reference			Reference
Smoking or alcohol	0.862 (0.215)	.55	(0.529, 1.404)	1.258 (0.305)	.34	(0.782, 2.024)	0.872 (0.201)	.55	(0.555, 1.370)	0.959 (0.192)	.84	(0.648, 1.421)
Both smoking and alcohol	0.946 (0.305)	.86	(0.503, 1.779)	1.039 (0.336)	.91	(0.551, 1.958)	0.952 (0.288)	.87	(0.526, 1.722)	0.702 (0.206)	.23	(0.396, 1.247)
Self-rated health
Very poor	Reference			Reference			Reference			Reference
Poor	3.642 (2.864)	.10	(0.780, 17.008)	2.323 (1.605)	.22	(0.600, 8.996)	3.072 (2.036)	.09	(0.838, 11.261)	1.160 (0.419)	.68	(0.572, 2.354)
Average	6.163 (4.717)	.02	(1.375, 27.627)	4.097 (2.702)	.03	(1.125, 14.924)	4.445 (2.855)	.02	(1.262, 15.655)	1.421 (0.494)	.31	(0.719, 2.808)
Good	6.952 (5.384)	.01	(1.524, 31.717)	3.757 (2.519)	.05	(1.010, 13.983)	4.614 (3.005)	.02	(1.287, 16.535)	1.595 (0.575)	.20	(0.787, 3.233)
Excellent	6.537 (5.172)	.02	(1.387, 30.820)	3.113 (2.146)	.10	(0.806, 12.024)	3.338 (2.236)	.07	(0.898, 12.405)	1.836 (0.707)	.11	(0.864, 3.905)
Vaccination belief
Negative belief	Reference			Reference			Reference			Reference
Positive belief	0.930 (0.204)	.74	(0.605, 1.429)	0.841 (0.183)	.43	(0.548, 1.289)	1.225 (0.259)	.34	(0.809, 1.854)	1.135 (0.206)	.48	(0.796, 1.619)
Trust in doctors
No trust	Reference			Reference			Reference			Reference
Trust	1.302 (0.366)	.35	(0.750, 2.258)	1.471 (0.431)	.19	(0.828, 2.614)	0.963 (0.236)	.88	(0.595, 1.557)	1.051 (0.224)	.81	(0.693, 1.596)
Perceived utility of online health information
Negative	Reference			Reference			Reference			Reference
Positive	0.660 (0.113)	.02	(0.472, 0.922)	0.876 (0.149)	.43	(0.628, 1.221)	1.053 (0.166)	.75	(0.773, 1.434)	1.123 (0.156)	.40	(0.856, 1.474)
Health insurance status
Uninsured	Reference			Reference			Reference			Reference
Basic health insurance	1.231 (0.553)	.64	(0.510, 2.968)	1.295 (0.630)	.60	(0.499, 3.362)	0.631 (0.218)	.18	(0.320, 1.243)	0.961 (0.312)	.90	(0.509, 1.814)
Commercial/supplementary/other	2.116 (1.218)	.19	(0.685, 6.539)	2.174 (1.310)	.20	(0.668, 7.079)	1.112 (0.531)	.83	(0.435, 2.837)	1.042 (0.482)	.93	(0.421, 2.580)
Hold two or more types	1.345 (0.673)	.55	(0.504, 3.589)	1.245 (0.662)	.68	(0.439, 3.528)	0.636 (0.256)	.26	(0.289, 1.400)	1.216 (0.455)	.60	(0.584, 2.532)
Internet usage
Not used	Reference			Reference			Reference			Reference
Used	1.410 (0.418)	.25	(0.789, 2.520)	1.396 (0.486)	.34	(0.706, 2.762)	1.356 (0.369)	.26	(0.796, 2.313)	1.336 (0.276)	.16	(0.892, 2.002)
Living in a smart-plan city
No	Reference			Reference			Reference			Reference
Yes	1.585 (0.267)	.01	(1.139, 2.205)	1.484 (0.250)	.02	(1.067, 2.065)	1.298 (0.200)	.09	(0.959, 1.756)	1.446 (0.195)	.01	(1.111, 1.883)
Constant	0.032 (0.034)	.001	(0.004, 0.252)	0.051 (0.053)	.004	(0.007, 0.390)	0.064 (0.057)	0.002	(0.011, 0.369)	0.061 (0.039)	<.001	(0.017, 0.212)
Observation	1831
R-squared	0.1009

In contrast, the presence of LTCs was not significantly associated with physical exercise frequency. Several demographic and psychosocial covariates, such as age and education were positively associated with daily exercise, while poorer self-rated health was linked to less frequent physical exercise.

Analysis of the impact of OHIS among patients with LTCs

Frequency of medical visits

Table 4 presents the multinomial logistic regression results for the interaction between OHIS and LTCs on the frequency of medical visits. After controlling for covariates, the interaction term (OHIS × LTCs) was statistically significant in the seldom (RRRs = 0.530, p = .04) and sometimes (RRRs = 0.393, p = .01) categories, but not in the often category (p = .59). The RRRs values were less than 1 in the first two categories, indicating lower relative risk ratios compared with the reference group (Frequency of medical visits = Never).

Table 4.

Analysis of the interaction effects between OHIS and LTCs on frequency of medical visits (multinomial logistic regression).

	Seldom			Sometimes			Often
	RRRs (Standard error)	P-values	95% CI	RRRs (Standard error)	P-values	95% CI	RRRs (Standard error)	P-values	95% CI
OHIS	2.149 (0.315)	<.001	(1.613, 2.863)	3.628 (0.755)	<.001	(2.413, 5.456)	1.226 (0.483)	.78	(0.486, 2.610)
LTCs	1.801 (0.405)	.01	(1.160, 2.797)	3.555 (0.929)	<.001	(2.131, 5.933)	4.835 (1.896)	<.001	(2.242, 10.429)
OHIS×LTCs	0.530 (0.162)	.04	(0.291, 0.966)	0.393 (0.134)	.01	(0.201, 0.769)	0.746 (0.403)	.59	(0.259, 2.149)
Control variables	Controlled
Constant	0.724 (0.476)	.62	(0.199, 2.625)	0.117 (0.088)	.004	(0.027, 0.510)	0.109 (0.121)	.046	(0.012, 0.966)
Observation	1831
R-squared	0.1109

Note. Though not reported here, the same variables of demographic characteristics, health conditions and behaviors, health awareness and social resources as shown in Table 2 were controlled for in this regression. See Supplement Table 3 for details.

Further examination shown in Table 5 indicates that OHIS had a significant positive association with the frequency of medical visits among individuals without LTCs, while the marginal effects of OHIS among those with LTCs were smaller. For individuals without LTCs, the probability of reporting Never medical visits decreased from 0.417 to 0.243 (p < .001), and the probability of reporting Sometimes increased from 0.139 to 0.259 (p < .001). Changes in the Seldom and Often categories were relatively smaller. Among individuals with LTCs, the probability of Never visits decreased from 0.249 to 0.220, while the probability of Sometimes increased from 0.248 to 0.306 (both p < .001).

Table 5.

Marginal effects of the interaction between OHIS and LTCs on frequency of medical visits.

		Margin	Standard error	P-value	95% CI
Never	Without LTCs × No OHIS	0.417	0.023	<.001	(0.373, 0.461)
	Without LTCs × OHIS	0.243	0.015	<.001	(0.213, 0.272)
	LTCs × No OHIS	0.249	0.032	<.001	(0.187, 0.312)
	LTCs × OHIS	0.220	0.032	<.001	(0.158, 0.283)
Seldom	Without LTCs × No OHIS	0.400	0.024	<.001	(0.354, 0.446)
	Without LTCs × OHIS	0.473	0.020	<.001	(0.434, 0.511)
	LTCs × No OHIS	0.402	0.035	<.001	(0.334, 0.471)
	LTCs × OHIS	0.401	0.036	<.001	(0.330, 0.472)
Sometimes	Without LTCs × No OHIS	0.139	0.017	<.001	(0.105, 0.173)
	Without LTCs × OHIS	0.259	0.018	<.001	(0.224, 0.295)
	LTCs × No OHIS	0.248	0.027	<.001	(0.196, 0.301)
	LTCs × OHIS	0.306	0.031	<.001	(0.245, 0.367)
Often	Without LTCs × No OHIS	0.044	0.012	<.001	(0.021, 0.068)
	Without LTCs × OHIS	0.025	0.007	<.001	(0.012, 0.039)
	LTCs × No OHIS	0.100	0.018	<.001	(0.066, 0.135)
	LTCs × OHIS	0.073	0.015	<.001	(0.044, 0.103)

These patterns are further illustrated in Figures 1–4. In Figure 1, which presents the predicted probability of Never medical visits, the line for the non-adopter group slopes downward as LTCs are present, whereas the line for the adopter group remains relatively flat, suggesting that the association between LTCs and lower frequencies of medical visits is more apparent among non-adopters. In contrast, Figure 2 shows a downward slope for the adopter group and a nearly flat pattern for the non-adopter group, indicating that LTCs status is more clearly related to variations in Seldom visits among OHIS adopters. In Figure 3, the separation between lines is again greater in the non-adopter group, implying a stronger association between LTCs and reporting Sometimes visits among those without OHIS. Figure 4 displays relatively parallel lines, suggesting minimal differences by LTCs status across both groups.

Figure 1.

Average marginal effects of OHIS with 95% CI (y=Never).

Figure 2.

Average marginal effects of OHIS with 95% CI (y=Seldom).

Figure 3.

Average marginal effects of OHIS with 95% CI (y=Sometimes).

Figure 4.

Average marginal effects of OHIS with 95% CI (y=Often).

Frequency of physical exercise

Table 6 presents the results of a multinomial logistic regression analysis examining physical exercise frequency based on OHIS-LTCs interaction groups, with the OHIS non-adopter without LTCs group serving as the reference group and relevant control variables included. Relative to the reference group, the OHIS non-adopter with LTCs group showed RRRs from 0.821 to 1.362 across categories, with no significant differences (all p>.05). The OHIS adopter without LTCs group had significantly higher RRRs in all categories (1.719-3.579, all p < .001). The OHIS adopter with LTCs group also had significantly higher RRRs (2.307-3.906, p≤.01).

Table 6.

Analysis of physical exercise frequency based on OHIS-LTCs interaction groups (multinomial logistic regression).

Groups	Frequency of physical exercise	RRRs (Standard error)	P-values	95% CI
OHIS non-adopter and without LTCs	A few times a year	Reference
	Several times a month
	Several times a week
	Daily
OHIS non-adopter but with LTCs	A few times a year	0.922 (0.292)	.80	(0.496, 1.715)
	Several times a month	1.362 (0.463)	.36	(0.699, 2.651)
	Several times a week	0.821 (0.237)	.49	(0.466, 1.444)
	Daily	1.242 (0.273)	.32	(0.807, 1.912)
OHIS adopter but without LTCs	A few times a year	2.344 (0.511)	<.001	(1.529, 3.594)
	Several times a month	3.579 (0.831)	<.001	(2.271, 5.642)
	Several times a week	2.193 (0.439)	<.001	(1.481, 3.246)
	Daily	1.719 (0.307)	<.001	(1.211, 2.441)
OHIS adopter and with LTCs	A few times a year	3.187 (1.027)	<.001	(1.694, 5.994)
	Several times a month	3.906 (1.303)	<.001	(2.031, 7.512)
	Several times a week	2.307 (0.697)	.01	(1.276, 4.173)
	Daily	2.597 (0.686)	<.001	(1.547, 4.359)
Control variables	Controlled
Observation	1831
R-squared	0.1013

Note. Though not reported here, the same variables of demographic characteristics, health conditions and behaviors, health awareness and social resources as shown in Table 3 were controlled for in this regression. See Supplement Table 4 for details.

Across groups, both OHIS adopter groups showed higher RRRs than the reference group in all categories, while the non-adopter with LTCs group had RRRs near 1 with no significant differences. The adopter with LTCs group displayed the highest RRRs in most categories.

Sensitivity analysis

To assess the robustness of the main findings, we conducted several sensitivity analyses. These involved estimating separate multinomial logistic regression models for Traditional Chinese Medicine visits and Western medicine visits, employing the original multi-category measure of OHIS frequency (with “never used” as the reference category, and levels including several times a year, several times a month, several times a week, once a day, and rarely used/several times a day), incorporating regional COVID-19 confirmed cases as an additional covariate, and excluding the health anxiety variable from the models. All models were re-estimated for both primary outcomes (frequency of medical visits and frequency of physical exercise), both with and without interaction terms for LTCs.

By employing the six-category OHIS, the results showed (Supplement Table 6) intermediate OHIS levels exhibiting the strongest effects such that “several times a month” yielded RRRs of 3.52 (p<.001) for sometimes visits, and the highest OHIS category (once or several times a day) showed weaker associations (RRRs=2.28, p<.001). A similar non-linear pattern emerged for physical exercise frequency (Supplement Table 7), where intermediate OHIS levels again produced the largest RRRs, particularly for lower exercise frequencies (“a few times a year” and “several times a month”). Interaction analyses (Supplement Tables 8 and 9) incorporating LTCs largely preserved this main-effect pattern of strongest associations at intermediate OHIS frequencies, although the effects varied according to long-term condition status. Overall, the results still show a positive correlation between OHIS and the dependent variables, consistent with the main model findings.

Although Chinese and Western medical services may differ conceptually, our sensitivity analyses indicated that their associations with OHIS were similar in both direction and significance (Supplementary Tables 10–15). Thus, aggregating visit frequencies provided a simplified yet empirically justified indicator of medical utilization in this context. And other results of these sensitivity analyses, which also confirmed the stability of the core findings, are presented in full in the Supplementary Materials (Supplementary Tables 16–20).

Discussion

This study examined how OHIS correlates with both healthcare service utilization (frequency of medical visits) and exercise-related self-management (frequency of physical exercise) in China during COVID-19. The results show that the OHIS is associated with higher medical visit frequency, particularly in the seldom and sometimes categories, with no significant association observed in the often category. OHIS is also positively correlated with higher levels of physical exercise frequency, ranging from several times a month to daily. The LTCs status itself is linked to increased medical visit frequency, especially in the sometimes and often categories, but does not show a significant overall association with physical exercise frequency. The positive association between OHIS and medical visit frequency is weaker in the group with LTCs (the interaction term RRRs<1). Regarding physical exercise, OHIS adopters with LTCs show the highest relative risk in most categories.

The finding of a positive association between OHIS and healthcare utilization aligns with some existing studies, which have observed a general positive correlation. However, this study further reveals heterogeneity in this association across different levels of healthcare visit frequencies and LTCs subgroups. The weaker association between OHIS and medical visits among individuals with LTCs may reflect the presence of already established patterns of healthcare utilization among these patients.^29,64 This is consistent with some studies showing that although individuals with LTCs use more healthcare services, their existing care routines satisfy much of their medical needs,⁶⁵ so additional information seeking has less impact on further medical visits.²⁹

This study observed a positive association between OHIS and physical exercise frequency, a pattern consistent with the viewpoint in existing literature that OHIS may support exercise-related self-management behaviors.^4,49 This could be interpreted as OHIS providing a potential informational basis for exercise-related self-management, driven by perceived benefits and social support among chronic disease patients.^4,49 This study further validates this finding. Interaction analysis indicates that OHIS adopters with LTCs show the highest relative risk ratios across most exercise categories (RRRs = 2.307-3.906), compared with non-adopters and those without LTCs. This suggests that online health information may have a more significant association with self-management behaviors (regular exercise) in the long-term patient population.

However, it is important to note that the unique context during the COVID-19 period may amplify the role of OHIS as a supplementary information channel. This aligns with findings from COVID-19-related studies, which have observed an increase in online information seeking behavior.^29,66 On the other hand, this situation, characterized by the China’s “avoid non-essential medical visits” policy, presents an ideal research situation, as the mechanisms linking OHIS to increased healthcare utilization remain insufficiently understood,^67,68 particularly given its potential to promote non-essential care.⁶⁹

And the results of the sensitivity analysis for the six-category OHIS may indicate that variations in OHIS are associated with different ways that people engage with health information resources. Moderate OHIS frequence might correspond to a more selective or purpose-driven use that helps individuals maintain focus without excessive exposure.⁷⁰ In contrast, very frequent seeking could reflect a broader or less targeted pattern of use that may involve exposure to more inconsistent information.⁷¹ Differences in motivation might also play a role, where frequent searching may be related to heightened health concerns rather than goal-oriented information use.^19,59 Limited uses might be linked to lower digital access or skills.⁷² Taken together, these possibilities point to a nuanced relationship between information-seeking behavior and health-related outcomes, which future research could examine through more detailed assessments of information quality and purpose.⁷³

These findings extend the application of the IMB model and the biopsychosocial model by providing empirical support for their utility in the context of digital health information seeking among patients with long-term conditions in resource-constrained settings. Accessible online information appears to strengthen motivation and behavioral skills for autonomous health behaviors, such as regular exercise, especially where formal healthcare access is limited.^12,74 The stronger effects at moderate levels of OHIS align with the IMB framework, indicating that information and motivation are most influential when engagement with digital content is purposeful and manageable.⁵⁹ However, it also shows the IMB model’s limitation in explaining the link between OHIS and medical visits, as behaviors directed by external factors and intricate psychological elements are not fully accounted for.⁷⁵ Moreover, the divergent ways in which the associations with medical visits and physical exercise differ by long-term condition status enrich the biopsychosocial model’s theoretical depth in digital interventions for the individual with LTCs. And it was verified that the long-tern condition status may amplify the supportive role of information on motivation and behavioral skills, particularly in the context of limited medical resources.⁴

Overall, these results underscore that OHIS is not a uniform process but is embedded in dynamic interactions among biological, psychological, and social factors.⁷⁶

Policy implications

Based on these findings, we propose the following practical implications as potential strategies for integrating OHIS into LTCs management. First, certified online portals offering verified self-care guidance should be promoted. Validated digital health information could be integrated into primary care through endorsement or advisory roles by family physician services and community health centers.

Second, quality control or fact-checking mechanisms for online health information could be established to help prevent misinformation, reduce health-related anxiety, and avoid wasteful use of healthcare services.

Third, eHealth literacy promotion could be institutionalized by embedding short digital navigation and information verification modules into LTCs education programs, with particular attention to older adults and lower-educated patients. These findings are particularly relevant for emerging economies facing economic burdens or financing challenges related to LTCs.

Limitations

This study has several limitations. First, due to the cross-sectional survey design, this study is correlational and does not attempt to establish causality. Second, residual confounding cannot be completely excluded. Unmeasured factors such as disease severity or digital skills may potentially influence both OHIS and health behaviors. The vaccination belief may be influenced by ideological beliefs, trust in health authorities, and other pandemic-related attitudes.^77,78 Third, key variables (OHIS, physical exercise, health anxiety) were self-reported and thus subject to recall or social desirability bias, although sensitivity analyses showed consistent results. Fourth, physical exercise captures only one facet of self-management. Furthermore, generalizability is constrained by China’s unique COVID-19 containment context and digital health infrastructure. Future research should employ longitudinal or multi-country designs and differentiate types of online health information (e.g., disease prevention, treatment recommendations, medication information) and presentation styles (e.g., loss vs. gain framing).⁵⁹

Conclusions

Using cross-sectional data from 1,831 adults in the 2021 China General Social Survey, this study examined associations between OHIS, healthcare utilization, and exercise-related self-management during the COVID-19 pandemic. OHIS was positively associated with more frequent medical visits and greater physical exercise. Notably, among individuals with LTCs, OHIS was associated with fewer medical visits but higher exercise engagement, consistent with a complementarity pathway toward self-management. Although causal interpretation is precluded by the cross-sectional design, the findings suggest that accessible online health information may support exercise-related self-management capacity, particularly in resource-constrained settings. From a policy perspective, integrating verified digital health content into primary care offers a low-cost strategy to mitigate the growing burden of LTCs in developing economies. Quality control mechanisms for online health information and eHealth literacy promotion play key complementary roles.

Supplemental material

Supplemental material - Online health information seeking, healthcare utilization, and exercise-related self-management among patients with long-term conditions in China during COVID-19

Supplemental material for Online health information seeking, healthcare utilization, and exercise-related self-management among patients with long-term conditions in China during COVID-19 by Yifan Jiang, Jinghua Zhang, Jianwei Wu in DIGITAL HEALTH

Footnotes

Acknowledgements

Data analyzed in this paper (article) were collected by the research project “Chinese General Social Survey (CGSS)” carried out by the National Survey Research Center (NSRC), Renmin University of China. We are grateful to the CGSS team for providing the data. In addition, we would like to express our gratitude to the editor and reviewers for their valuable and constructive suggestions, which have greatly helped us improve the manuscript.

ORCID iDs

Yifan Jiang

Jinghua Zhang

Jianwei Wu

Ethical considerations

The Research Ethics Committee of School of Business, Macau University of Science and Technology (MUST) has examined and issued Ethical Approval. The relevant Ref. No. is MSB-202529.

Author Contributions

YFJ: Conceptualization; Data curation; Formal analysis; Methodology; Software; Writing – original draft; Writing – review & editing.

JHZ: Conceptualization; Formal analysis; Funding acquisition; Investigation; Project administration; Supervision; Writing – review & editing.

JWW: Project administration; Resources; Supervision; Validation; Visualization; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: JHZ and YFJ were supported by Macau University of Science and Technology Foundation, Faculty Research Grant (FRG-25-077-MSB). The funding parties have no role in the conduct or direction of the research. The researchers maintain full independence and autonomy in designing and carrying out the study, without interference or influence from the funding source. The research findings and conclusions were based solely on the scientific merit of the data collected, without any external input or oversight from the parties providing financial support for the project.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Declaration of generative AI

AI tools were not used in any aspect of the development of the manuscript. Large language models, specifically Gemini-3 and GPT-5, were utilized solely to help polish the English language expressions in the manuscript.

Supplemental material

Supplemental material for this article is available online.

Appendix

References

WHO . Noncommunicable diseases. https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases (2024, accessed 5/10/2025).

Khunti

Chudasama

Gregg

, et al. Diabetes and multiple long-term conditions: a review of our current global health challenge. Diabetes Care 2023; 46: 2092–2101. https://doi.org/10.2337/dci23-0035

Sloggett

. The forgotten majority? https://www.futurehealth-research.com/site/wp-content/uploads/2023/12/Long-Term-Conditions-Report-FINAL-DECEMBER-2023.pdf (2023).

Zhao

Song

. Online health information seeking among patients with chronic conditions: integrating the health belief model and social support theory. Journal of Medical Internet Research 2022; 24: e42447. https://doi.org/10.2196/42447

Lewinski

Walsh

Rushton

, et al. Telehealth for the longitudinal management of chronic conditions: systematic review. Journal of medical Internet research 2022; 24: e37100. https://doi.org/10.2196/37100

Allegrante

Wells

Peterson

. Interventions to support behavioral self-management of chronic diseases. Annual review of public health 2019; 40: 127–146. https://doi.org/10.1146/annurev-publhealth-040218-044008

WHO . Improving patient education: a new guide for policy-makers and health professionals to support self-management of chronic conditions. https://www.who.int/europe/news/item/14-11-2023-improving-patient-education--a-new-guide-for-policy-makers-and-health-professionals-to-support-self-management-of-chronic-conditions (2024, accessed 05/10/2025).

Xue

Cai

Liu

, et al. Trends and regional variations in chronic diseases and their risk factors in China: an observational study based on National Health Service Surveys. International Journal for Equity in Health 2023; 22: 120. https://doi.org/10.1186/s12939-023-01910-w

Cong

Liu

, et al. Family doctor contract services for persons with disabilities in China. Disability & Society 2026; 41: 587–592. https://doi.org/10.1080/09687599.2025.2498416

10.

Zhang

Peng

, et al. Barriers and facilitators of family doctor contract services in caring for disabled older adults in Beijing, China: a mixed methods study. BMJ open 2023; 13: e070130. https://doi.org/10.1136/bmjopen-2022-070130

11.

Meng

. Impacts of the internet on health inequality and healthcare access: a cross-country study. Frontiers in Public Health 2022; 10: 935608. https://doi.org/10.3389/fpubh.2022.935608

12.

Liu

Yang

Cheng

, et al. Online health information seeking, eHealth literacy, and health behaviors among Chinese internet users: cross-sectional survey study. Journal of Medical Internet Research 2024; 26: e54135. https://doi.org/10.2196/54135

13.

Madrigal

Escoffery

. Electronic health behaviors among US adults with chronic disease: cross-sectional survey. Journal of medical Internet research 2019; 21: e11240. https://doi.org/10.2196/11240

14.

Prestin

Vieux

Chou

W-yS

. Is online health activity alive and well or flatlining? Findings from 10 years of the Health Information National Trends Survey. Journal of health communication 2015; 20: 790–798. https://doi.org/10.1080/10810730.2015.1018590

15.

Chae

. How we use the Internet matters for health: The relationship between various online health-related activities and preventive dietary behaviors. Health Informatics Journal 2019; 25: 973–983. https://doi.org/10.1177/1460458217735675

16.

Liu

. The effects of older Chinese adults’ online behaviors on their health habits and health status. Digital Health 2025; 11: 20552076251359130. https://doi.org/10.1177/20552076251359130

17.

Mitsutake

Takahashi

Otsuki

, et al. Chronic diseases and sociodemographic characteristics associated with online health information seeking and using social networking sites: nationally representative cross-sectional survey in Japan. Journal of Medical Internet Research 2023; 25: e44741. https://doi.org/10.2196/44741

18.

Kbaier

Kane

McJury

, et al. Prevalence of health misinformation on social media—challenges and mitigation before, during, and beyond the covid-19 pandemic: Scoping literature review. Journal of Medical Internet Research 2024; 26: e38786. https://doi.org/10.2196/38786

19.

McMullan

Berle

Arnáez

, et al. The relationships between health anxiety, online health information seeking, and cyberchondria: Systematic review and meta-analysis. Journal of affective disorders 2019; 245: 270–278. https://doi.org/10.1016/j.jad.2018.11.037

20.

Brown

Skelly

Chew‐Graham

. Online health research and health anxiety: A systematic review and conceptual integration. Clinical psychology: Science and practice 2020; 27: e12299. https://doi.org/10.1111/cpsp.12299

21.

Zheng

Chen

Jiang

, et al. How does health information seeking from different online sources trigger cyberchondria? The roles of online information overload and information trust. Information Processing & Management 2023; 60: 103364. https://doi.org/10.1016/j.ipm.2023.103364

22.

Hannah

Marie

Olaf

, et al. The global economic burden of health anxiety/hypochondriasis-a systematic review. BMC Public Health 2023; 23: 2237. https://doi.org/10.1186/s12889-023-17159-5

23.

Van Wilder

Pype

Mertens

, et al. Living with a chronic disease: insights from patients with a low socioeconomic status. BMC family practice 2021; 22: 1–11. https://doi.org/10.1186/s12875-021-01578-7

24.

Zhu

Zhao

Yun

, et al. Research on health information avoidance behavior and influencing factors of cancer patients—an empirical analysis based on structural equation modeling. BMC Public Health 2024; 24: 3617. https://doi.org/10.1186/s12889-024-21113-4

25.

Foust

Taber

. Information avoidance: Past perspectives and future directions. Perspectives on Psychological Science 2025; 20: 241–263. https://doi.org/10.1177/17456916231197668

26.

Meng

Song

, et al. Information avoidance and medical screening: A field experiment in China. Management Science 2021; 67: 4252–4272. https://doi.org/10.1287/mnsc.2020.3723

27.

Suziedelyte

. How does searching for health information on the Internet affect individuals' demand for health care services? Social science & medicine 2012; 75: 1828–1835. https://doi.org/10.1016/j.socscimed.2012.07.022

28.

Yigzaw

Wynn

Marco-Ruiz

, et al. The association between health information seeking on the internet and physician visits (the seventh tromsø study-part 4): population-based questionnaire study. Journal of medical Internet research 2020; 22: e13120. https://doi.org/10.2196/13120

29.

Di Novi

Kovacic

Orso

. Online health information seeking behavior, healthcare access, and health status during exceptional times. Journal of Economic Behavior & Organization 2024; 220: 675–690. https://doi.org/10.1016/j.jebo.2024.02.032

30.

Suenaga

Vicente

. Online and offline health information seeking and the demand for physician services. The European Journal of Health Economics 2022; 23: 337–356. https://doi.org/10.1007/s10198-021-01352-7

31.

Zhong

Cao

Xue

. The paradox of convenience: how information overload in mHealth apps leads to medical service overuse. Frontiers in Public Health 2024; 12: 1408998. https://doi.org/10.3389/fpubh.2024.1408998

32.

Liu

Liang

. The dynamic COVID-zero strategy in China. China CDC weekly 2022; 4: 74–75. https://doi.org/10.46234/ccdcw2022.015

33.

WHO . Infodemic. https://www.who.int/health-topics/infodemic#tab=tab_1 (2020, accessed 5/12/2025).

34.

Jungmann

Witthöft

. Health anxiety, cyberchondria, and coping in the current COVID-19 pandemic: Which factors are related to coronavirus anxiety? Journal of anxiety disorders 2020; 73: 102239. https://doi.org/10.1016/j.janxdis.2020.102239

35.

Tyrer

. COVID‐19 health anxiety. World Psychiatry 2020; 19: 307–308. https://doi.org/10.1002/wps.20798

36.

Fang

Nie

Penny

. Transmission dynamics of the COVID‐19 outbreak and effectiveness of government interventions: A data‐driven analysis. Journal of medical virology 2020; 92: 645–659. https://doi.org/10.1002/jmv.25750

37.

Wilder-Smith

Freedman

. Isolation, quarantine, social distancing and community containment: pivotal role for old-style public health measures in the novel coronavirus (2019-nCoV) outbreak. Journal of travel medicine 2020; 27: taaa020. https://doi.org/10.1093/jtm/taaa020

38.

NHS . Notice of the General Office of the National Health Commission on Strengthening Medical Service Management During the Pandemic to Meet Essential Healthcare Needs. https://www.gov.cn/zhengce/zhengceku/2020-02/17/content_5480176.htm (2020, accessed 5/22/2025).

39.

van der Miesen

Raaijmakers

van de Grift

. “You have to wait a little longer”: Transgender (mental) health at risk as a consequence of deferring gender-affirming treatments during COVID-19. Archives of sexual behavior 2020; 49: 1395–1399. https://doi.org/10.1007/s10508-020-01754-3

40.

Liu

Chen

Liu

, et al. Comorbid chronic diseases are strongly correlated with disease severity among COVID-19 patients: a systematic review and meta-analysis. Aging and disease 2020; 11: 668–678. https://doi.org/10.14336/AD.2020.0502

41.

Kendzerska

Zhu

Gershon

, et al. The effects of the health system response to the COVID-19 pandemic on chronic disease management: a narrative review. Risk management and healthcare policy 2021; 14: 575–584. https://doi.org/10.2147/RMHP.S293471

42.

Saqib

MAN

Siddiqui

Qasim

, et al. Effect of COVID-19 lockdown on patients with chronic diseases. Diabetes & Metabolic Syndrome: Clinical Research & Reviews 2020; 14: 1621–1623. https://doi.org/10.1016/j.dsx.2020.08.028

43.

Engel

. The clinical application of the biopsychosocial model. The Journal of medicine and philosophy 1981; 6: 101–124. https://doi.org/10.1093/jmp/6.2.101

44.

Fisher

W. A.

Fisher

J. D

. The information-motivation-behavioral skills model: A general social psychological approach to understanding and promoting health behavior. In J. Suls & K. A. Wallston (Eds.), Social psychological foundations of health and illness (pp. 82–106). Wiley-Blackwell.

45.

Hartch

Dietrich

Lancaster

, et al. Effects of a medication adherence app among medically underserved adults with chronic illness: a randomized controlled trial. Journal of Behavioral Medicine 2024; 47: 389–404. https://doi.org/10.1007/s10865-023-00446-2

46.

Yang

Luo

, et al. Unpacking cyberchondria: The roles of online health information seeking, health information overload, and health misperceptions. Telematics and Informatics 2025; 97: 102225. https://doi.org/10.1016/j.tele.2024.102225

47.

Peng

. How online searches fuel health anxiety: Investigating the link between health-related searches, health anxiety, and future intention. Computers in Human Behavior 2022; 136: 107384. https://doi.org/10.1016/j.chb.2022.107384

48.

Knoop

Jones

Norton

, et al. Biopsychosocial factors associated with distress in people with suspected postural orthostatic tachycardia syndrome (POTS): A longitudinal regression and correlation study/a quantitative study. Journal of Psychosomatic Research 2025: 112499.

49.

Liu

Huang

Xiao

, et al. Online Health Information Seeking: Implications for Self-Management in Hypertension. Risk Management and Healthcare Policy 2025; 18: 2951–2965. https://doi.org/10.2147/RMHP.S539905

50.

Jabkowski

Cichocki

Kołczyńska

. Multi-project assessments of sample quality in cross-national surveys: the role of weights in applying external and internal measures of sample bias. Journal of Survey Statistics and Methodology 2023; 11: 316–339. https://doi.org/10.1093/jssam/smab027

51.

Bian

. The Chinese general social survey (2003-8) sample designs and data evaluation. Chinese Sociological Review 2012; 45: 70–97. https://doi.org/10.2753/csa2162-0555450104

52.

Wang

Zhou

, et al. Useful or not? The discussion of traditional Chinese medicine to treat COVID-19 on a Chinese social networking site. BMJ Global Health 2024; 9: e014398. https://doi.org/10.1136/bmjgh-2023-014398

53.

Liao

Cheng

, et al. The mediating role of physical activity and health status between a health-supportive environment and well-being: a cross-sectional study. Frontiers in Public Health 2023; 11: 1233970. https://doi.org/10.3389/fpubh.2023.1233970

54.

Lee

Ghang

. Healthcare utilization among foreign beneficiaries of the National Health Insurance program in Korea. BMC Health Services Research 2024; 24: 1312. https://doi.org/10.1186/s12913-024-11698-2

55.

Schmalbach

Franke

Häuser

, et al. Attachment styles and healthcare utilization: exploring the role of the patient-doctor relationship. BMC Health Services Research 2024; 24: 63. https://doi.org/10.1186/s12913-023-10484-w

56.

Nielsen

Seibæk

Pedersen

, et al. Health literacy and self-management of people living with chronic obstructive pulmonary disease or type 2 diabetes in Greenland: a qualitative interpretive description study. BMC Public Health 2025; 25: 3106. https://doi.org/10.1186/s12889-025-24431-3

57.

Hou

. The negative association between health status and digital health literacy among Chinese youth: a cross-sectional study on Chinese general social survey. Frontiers in Public Health 2025; 13: 1682986. https://doi.org/10.3389/fpubh.2025.1682986

58.

Budge

Taylor

Curtis

. Support for living well with long‐term conditions: How people manage. Journal of clinical nursing 2021; 30: 475–487. https://doi.org/10.1111/jocn.15560

59.

Wang

Shi

Kong

. Online health information seeking: a review and meta-analysis. Health Communication 2021; 36: 1163–1175. https://doi.org/10.1080/10410236.2020.1748829

60.

Williams

. Generalized ordered logit/partial proportional odds models for ordinal dependent variables. The stata journal 2006; 6: 58–82. https://doi.org/10.1177/1536867x0600600104

61.

Rysavy

Bobashev

, et al. Comparing methods for risk prediction of multicategory outcomes: dichotomized logistic regression vs. multinomial logit regression. BMC Medical Research Methodology 2024; 24: 261. https://doi.org/10.1186/s12874-024-02389-x

62.

Agresti

Kateri

. Categorical data analysis. In: International encyclopedia of statistical science. Springer, 2025, pp. 408–411.

63.

Tutz

. Ordinal regression: A review and a taxonomy of models. Wiley Interdisciplinary Reviews: Computational Statistics 2022; 14: e1545. https://doi.org/10.1002/wics.1545

64.

Zhai

, et al. Associations Among Online Health Information Seeking Behavior, Online Health Information Perception, and Health Service Utilization: Cross-Sectional Study. Journal of Medical Internet Research 2025; 27: e66683. https://doi.org/10.2196/66683

65.

Gaglioti

Baltrus

, et al. Interpersonal primary care continuity for chronic conditions is associated with fewer hospitalizations and emergency department visits among Medicaid enrollees. The Journal of the American Board of Family Medicine 2023; 36: 303–312. https://doi.org/10.3122/jabfm.2022.220015R1

66.

Van Kessel

Kyriopoulos

Wong

BLH

, et al. The effect of the COVID-19 pandemic on digital health–seeking behavior: big data interrupted time-series analysis of google trends. Journal of Medical Internet Research 2023; 25: e42401. https://doi.org/10.2196/42401

67.

Arruda

Ayoub

Nunes

, et al. Consulting “Dr. Google”: how the digital search for internet health information influences doctor-patient relationship. Cadernos de Saúde Pública 2025; 41: e00153623. https://doi.org/10.1590/0102-311XEN153623

68.

Kobryn

Duplaga

. Cyberchondria severity and utilization of health services in Polish society: a cross-sectional study. BMC Public Health 2024; 24: 902. https://doi.org/10.1186/s12889-024-18399-9

69.

Daraz

Morrow

Ponce

, et al. Can patients trust online health information? A meta-narrative systematic review addressing the quality of health information on the internet. Journal of general internal medicine 2019; 34: 1884–1891. https://doi.org/10.1007/s11606-019-05109-0

70.

Chen

Yang

Xia

, et al. Patterns and associated factors of online health information seeking behaviors among young women diagnosed with breast cancer in China. Asia-Pacific Journal of Oncology Nursing 2025; 12: 100700. https://doi.org/10.1016/j.apjon.2025.100700

71.

Xiong

Zhang

, et al. Frequency of online health information seeking and types of information sought among the general Chinese population: cross-sectional study. Journal of Medical Internet Research 2021; 23: e30855. https://doi.org/10.2196/30855

72.

Guo

Zhao

Guo

, et al. Socioeconomic disparities in eHealth literacy and preventive behaviors during the COVID-19 pandemic in Hong Kong: cross-sectional study. Journal of medical Internet research 2021; 23: e24577. https://doi.org/10.2196/24577

73.

Lim

Dunn

Lim

, et al. Association between online health information-seeking and medication adherence: a systematic review and meta-analysis. Digital health 2022; 8: 20552076221097784. https://doi.org/10.1177/20552076221097784

74.

Zhou

Zhao

Xiao

, et al. The impact of online health community engagement on lifestyle changes: A serially mediated model. Frontiers in public health 2022; 10: 987331. https://doi.org/10.3389/fpubh.2022.987331

75.

Jung

Song

. Testing the mediating effect of risk perception on the relationship between e-health literacy and infectious disease prevention behaviors among young adults based on the IMB model. Frontiers in Public Health 2025; 13: 1728099. https://doi.org/10.3389/fpubh.2025.1728099

76.

Zhao

Yang

Miao

, et al. Factors Associated With the Level of Trust in Health Information Robots Among the General Population From a Socioecological Model Perspective: Network Analysis. Journal of Medical Internet Research 2025; 27: e68299. https://doi.org/10.2196/68299

77.

Chen

Wang

, et al. Higher collective responsibility, higher COVID-19 vaccine uptake, and interaction with vaccine attitude: Results from propensity score matching. Vaccines 2022; 10: 1295. https://doi.org/10.3390/vaccines10081295

78.

Chen

, et al. Predicting COVID-19 vaccination timing by integrating the theory of planned behavior and the diffusion of innovations: a cross-sectional survey in Macao, China. Journal of Thoracic Disease 2025; 17: 2813–2826. https://doi.org/10.21037/jtd-24-1313

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

2.26 MB