A cross-sectional online survey was conducted in mainland China. The study protocol was approved by the Research Ethics Committee of Harbin Medical University (IRB code HMUIRB20200004). The survey was anonymous. Participant information sheet was provided and implied informed consent was required from each participant prior to proceeding to the survey.

Study participants were recruited through the online survey platform Wenjuanxing (www.wjx.cn). It has reach to the largest pool of survey participants in mainland China, covering all regions: more than one million questionnaire responses are recorded by Wenjuanxing every day. Eligible participants in this study were the adults over 18 years of age. They were identified randomly through an automation process embedded in the Wenjuanxing sampling services. The identity of the invited participants remained anonymous and unknown to the research team.

Data were collected in March 2021. Only one submission per IP address was allowed. The survey closed when returned questionnaires reached 4,383. To ensure quality of the data, the returned questionnaires containing logical errors (contradictory answers) and those that were completed within 10 min were excluded. Our pilot test showed that at least 10 min would be needed to read through the questionnaire. This resulted in a final sample size of 4,325 (98.75% of the returned questionnaires). Of those, 1,317 (30.45%) reported health problems and intention to seek medical care.

The questionnaire development was informed by the existing literature (29) and was adapted to the specific context of COVID-19 in line with the relevant guidelines issued by the WHO (30), the National Health Commission and the national CDC in China (31).

Delay in seeking health care was the main interest of the study. Participants were asked to report their self-assessment of health and intention to seek health care over the past month, when the seven-day rolling average of daily new confirmed COVID-19 cases ranged from 7.57 to 56.71 (32). For those who intended to seek health care, their experiences in obtaining the needed care were further investigated through a series of questions, which included whether they ‘delayed care due to concerns related to COVID-19' (1 = “yes”; 0 = “no/unsure”), for what condition (cardiovascular and/or cerebrovascular diseases, digestive diseases, bone diseases; respiratory disease, eye nose throat diseases, diabetes mellitus, tumor, accident and injury, and others), through what service (medical consultation, emergency treatment (care for immediate life-threatening conditions), obtainment of medicines, follow-up examination, hospital admission, surgical procedure, and others), and from which provider (local provincial/municipal public hospitals, primary healthcare network, cross-provincial/municipal public hospitals, private clinics or private hospitals). In addition, they were asked to identify one or more reasons for the delay, if applied, from the following list: fear of infection; discouragement from relatives and friends; difficulties with online appointments; long waiting time in facilities; complex service procedure; facility closure; denied access to facilities; transfer to infection/fever clinics; movement restrictions; community lockdown; complex referral procedure; and others.

The study participants who reported delay in health care were also asked to identify one or more consequences they anticipated (including disrupted medication, slow recovery, complications, missed optimal timing of treatment, deterioration of illness conditions, dissatisfaction with care provision, increased costs, increased mental burden on family, and others) and how they coped with the delay (including use of Internet-based medical services, self-treatment at home, family/friend help, telemedicine, government assistance, and others).

The selection of independent variables was guided by the Andersen's service utilization model (29), which categorizes predictors of health service utilization into predisposing, enabling, and need factors. In this study, gender (male vs. female), age ( ≤ 30, 31–59, ≥60 years), and marital status (married vs. others) were deemed as predisposing factor, while residency (urban vs. rural), risk level of regions (high vs low), educational attainment (with or without a university degree), personal income (≥average (5,000) vs. < 5,000 Yuan per month (33)), health insurance coverage (yes vs. no), pregnancy or co-habitant with a pregnant woman (yes vs. no), and using Internet medical services (yes vs. no) measured enabling factor. In China, more than 50 accumulative active cases of COVID-19 in a municipality over a period of 14 days would be classified as high risk (7).

Need factor was measured by COVID-19 risk perception and chronic conditions (yes vs. no). The measurement of risk perception followed the definition of Bauer from Harvard (34), considering people's cognition, feeling, and comprehension of the risk characteristics, not the actual risk (35). Adams and Smith (36) pointed out that individual risk perception is closely associated the severity and controllability of the risk. Risk perception is a pivotal determinant of care seeking behaviors (37). In this study, a three-dimensional scale was adopted to measure COVID-19 risk perception, covering perceived susceptibility to COVID-19 infection, perceived severity of the consequences of COVID-19, and perceived controllability of COVID-19 outbreaks. Each dimension contains three items rated on a six-point Likert scale ranging from 1 (strongly disagree) to 6 (strongly agree). A summed average score was calculated for each dimension, with 1–3 indicating a low level and 4–6 indicating a high level of risk perception. The risk perception scale has been validated in a previous study (38). Good internal consistency (Cronbach's α = 0.824) and construct validity (GFI = 0.982, AGFI = 0.961, IFI = 0.972, CFI = 0.972, RMSEA = 0.062 in confirmatory factor analysis) of the scale were also evident in this study. Chronic conditions were defined as a general term for the diagnosed diseases with an insidious onset and prolonged course (39), which include cardiovascular diseases, cerebrovascular diseases, diabetes, and others (40).

The percentage distributions of the study participants with different characteristics were described and compared between those living in high and low risk regions using Chi-square tests. A multivariate logistic regression model was then established to determine the significant predictors of delay in seeking health care after adjustment for variations in other variables. The reasons of delay and perceived consequences were described and ranked in order using percentage distributions.

All data analyses were performed using SPSS 23.0. A two-sided p value of < 0.05 was considered as statistically significant.

Of the 1,317 study participants who reported a need to seek health care, 832 (63.2%) lived in high risk regions over the study period. Over half were female (54.1%), married (54.5%), obtained a university qualification (52.2%), and had no chronic conditions (55.4%) at the time of the survey. The vast majority were younger than 60 years (95.2%), resided in an urban area (67.1%), earned < 5,000 Yuan per month (61.7%), had health insurance coverage (93.5%), and were not pregnant or living with a pregnant woman (95.9%). Although 81.9% of respondents perceived high severity of COVID-19, 70.1% perceived high levels of controllability, and 87.9% perceived low levels of susceptibility. Compared with the respondents from a region with low mobility restrictions, those experiencing high mobility restrictions were older (p < 0.001), and were more likely to be married (p < 0.001), obtained no university qualifications (p = 0.01), earned a low level of income (p = 0.012), had no health insurance coverage (p < 0.001), lived with chronic conditions (p = 0.01), and perceived higher levels of susceptibility (p < 0.001) and lower levels of controllability (p = 0.012) (Table 1).

Overall, 31.4% of the study participants who had a care need experienced delay in seeking health care: 35.8% in those residing in a high risk region compared with 23.7% in those from low risk regions (p < 0.001). However, living in high risk regions (AOR = 1.736 [95% CI 1.307–2.334]) was not the only predictor of delay in seeking health care. An age between 31 and 59 years (AOR = 1.535 [95% CI 1.132–2.246]), lower levels of perceived controllability (AOR = 1.591 [95% CI 1.187–2.131]), living with chronic conditions (AOR = 2.008 [95% CI 1.544–2.611]), pregnancy or co-habitant with a pregnant woman (AOR = 2.115 [95% CI 1.154–3.874]), and access to Internet-based medical services (AOR = 2.529 [95% CI 1.960–3.265]) were also associated with delay in seeking health care according to the results of the multivariate modeling. The associations of delay in seeking health care with middle age, marriage, urban residency, higher personal income, and absence of health insurance coverage became statistically insignificant after adjustment for variations of other variables. The multivariate model explained 19.2% of variance (R²) in delay of seeking health care (Table 2).

Table 3 shows that the types of care being delayed appeared to be consistent between those from the high and low risk regions, despite some differences in the percentage distributions. Medical consultations (38.7%), emergency treatment (18.2%), and obtainment of medicines (16.5%) were the top three types of delayed care. Eye, nose, and throat diseases (23.2%) and cardiovascular and cerebrovascular diseases (20.8%) were the top two conditions relating to the delayed care, followed by digestive diseases (10.1%) in those living in a region with high risk and bone diseases (12.2%) and respiratory diseases (12.2%) in those living in a region with low risk. Most of the delayed services were planned to be obtained from local provincial/municipal hospitals (68.0%) and primary health care networks (17.2%). Self-treatment at home was the most likely coping strategy (34.9%), followed by Internet-based medical care (29.2%) and family/friend help (24.0%) (Table 3).

Figure 1 shows that the top three reasons of delay in seeking health care were fear of infection (52.01−57.39%), complex service procedure (40.94%~43.48%), and long waiting time in facilities (21.74−22.15%) according to the reports of the respondents from both regions with high and low COVID-19 risk. Of the other reasons, those from the regions with high risk were more likely to report facility closure (20.13 vs. 14.78%), but less likely to report transference to infection/fever clinics (18.26 vs. 10.07%) as a reason of the delay compared with their counterparts from the regions with low risk (Figure 1).

Figure 2 shows that the perceived consequences of delayed care followed the same pattern between those living in the regions with high and low risk. Slow recovery (43.29−50.43%) was the top concern, which was followed by disruptions in medication (39.13−42.62%), psychological distress (37.92−38.26%), missed optimal timing of treatment (27.85−33.04%), and deterioration of illness conditions (14.78%−16.44%) (Figure 2).

This study adopted a cross-sectional design, which does not allow us to compare the levels of delay in seeking health care before and after the outbreak of COVID-19. No causal conclusions can be established either. The population was not stratified for sampling although a simple random sampling strategy was adopted through the Wenjuanxing platform. We intended to obtain a maximal sample size without calculating the statistical power. The final sample size far exceeds the requirements of a statistical power of 0.8, with an α of 0.05. The final study sample was biased toward those residing in the regions with high risk and those younger than 60 years. The measurement of delay is also subject to recall bias. Attempts to generalize the results to the entire population in China need to be cautious.