Decline of onset-to-diagnosis interval and its impacts on clinical outcome of COVID-19 in China: a nation-wide observational study

Since COVID-19 was included in the list of Chinese notifiable infectious disease on January 21, 2020, each COVID-19 case was required to report to China CDC though the National Notifiable Infectious Diseases Surveillance System of China (NNIDSS) [28]. Individual data, including demographical information, exposure history, clinical manifestations and signs, physical examinations [e.g., chest X-ray or computed tomography (CT) scanning], clinical complications, and underlying condition were collected from the medical records.

This retrospective study was performed based on the data of confirmed COVID-19 cases reported through NNIDSS from February 5, 2020 to October 8, 2020. According to the National Health Commission of the People's Republic of China “Diagnosis and Treatment Program for Novel Coronavirus Infected Pneumonia” [29], all COVID-19 cases were confirmed by real-time polymerase chain reaction (PCR) and clinical manifestations. In addition, in the early stage of the epidemic, medical imaging examinations such as CT and X-rays were also used to confirm cases in Wuhan. We excluded people younger than 40 years old due to their low mortality and low severe rates. In addition, asymptomatic cases and imported cases from abroad were also excluded from the study.

This study used death and clinical severity as main study outcomes. Severe and mild cases were classified by the “Diagnosis and Treatment Scheme of New Coronavirus Infected Pneumonia” [29].

National Health Commission of the People’s Republic of China decided to collect and analyze the COVID-19 case data to control the outbreak. This study was approved by the ethical review committee of the Chinese Center for Disease Control and Prevention (2020-026). Data were de-identified, and informed consent was waived by the ethical review committee of the Chinese Center for Disease Control and Prevention.

For confirmed patients with COVID-19, the onset-to-diagnosis interval (ODI) was defined as the interval in days from symptom onset to laboratory diagnosis. In China, patients with positive nucleic acid tests were defined as confirmed cases after they had clinical symptoms or signs, so the ODI of all cases was greater than or equal to 0.

The median and interquartile range (IQR) were used to describe the ODIs, which were compared among sex, and age groups (40‒59 years, 60‒69 years, and ≥ 70 years) by Wilcoxon rank-sum test or Kruskal–Wallis rank-sum test. The ODI, severe rate (SR) and CFR in all the cases were delineated and compared between two regions (Wuhan vs. outside Wuhan).

The factors potentially associated with ODI were analyzed using multivariate linear regression model, which had accounted for sex, age and region. The profile of SR and CFR in response to the ODI was explored by Join-Point regression (JPR) model to identify the turning points of ODIs at which the significant change of SR and CFR could be marked. Given that CFRs and SRs followed abnormal distribution, natural log-linear model was used to estimate the linear trends of annual percentage change (APC) as recommended. Under a log-linear model, rates change at a constant percentage by each day of the interval, was calculated as a fixed APC. Positive or negative value of APC with statistical significance indicated an increasing or decreasing trend respectively, while non-significant (P ≥ 0.05) APC was considered as stable trend.

Logistic regression model was applied to evaluate the effects of ODI on the severe or fatal outcome. For the categorical grouping, the ODI within 19 days was subgrouped by a two-day interval, i.e., 0‒1, 2‒3, 4‒5…, and 18‒19. A series of logistic regression models were applied to evaluate the impact of different ODIs with the shortest interval of 0–1 day used as reference. All analysis was performed with sex, age and geographic region included, based on which the adjusted odds ratio (OR) and 95% confidence interval (CI) were estimated. In addition, attributable fraction (AF) of ODI to severe or fatal outcome was estimated by age, sex and region.

We estimated the possible overall numbers of severe cases and deaths under predesigned 16 scenarios, in which if all patients were diagnosed within 0, 1, 2, …, and ≥ 15 days after symptom onset. Based on the real SRs and CFRs reported under different ODIs from 0 to ≥ 15 days in 12 patient subgroups classified by age groups (40‒59 years, 60‒69 years, and ≥ 70 years), sex (male and female), and regions (Wuhan and outside Wuhan), we projected the numbers of severe or deceased cases for each subgroups and respectively for each of the predesigned 16 scenarios, and then the overall number of severe or deceased cases were estimated by summating the numbers of 12 subgroups for each of 16 scenarios.

We profiled the ODIs in terms of age and sex for Wuhan and outside Wuhan separately. The overall median ODI was 4 days (IQR 2‒7), while Wuhan had longer median ODIs than outside Wuhan (Table 1, Additional file 1: Fig. S2). In general, the ODI appeared to be slightly longer in female cases and cases aged 60‒69 years, and notably longer in Wuhan and among both severe cases than their counterparts (all P < 0.001, Table 1). However, fatal cases did not appear to have a longer ODI relative to non-death cases.

Multivariate linear regression model further revealed significantly longer ODIs in female, cases aged 60‒69 years and cases in Wuhan (Additional file 1: Table S1). Subgroup analysis showed that the pattern of a longer ODI in sex was consistently seen in both regions, but cases aged 60‒69 years did not have a longer ODI in Wuhan.

As a whole the SR increased from 7.0% to 20.0% and then decreased to 8.0%, corresponding to the increase of ODI from 0 to 20 days, with a turning point observed at Day 10 of the ODI. Before the turning point, the APC of SR was 8.61% (P < 0.05), but after that it was -6.66% (P < 0.05) (Fig. 2A).

With the increase of ODI, earlier appearing of turning points and higher APCs of SR were found in patients of males over females (10 days vs. 13 days for the turning points, and 10.50% vs. 8.25% for APCs), in patients aged 40‒59 years, 60‒69 years, and aged ≥ 70 years (10 days, 9 days and 13 days for the turning points, 13.47%, 14.90% and 4.85% for APCs), and outside Wuhan over Wuhan (10 days vs. 13 days for the turning points, 10.64% vs. 5.53% for APCs), respectively (Fig. 2B‒D). Given the same ODI, older age and from Wuhan both led to a higher SR (Fig. 2C‒D). The turning points of ODI for those aged < 70 years were significantly longer outside Wuhan than in Wuhan, while it was comparable and those aged ≥ 70 years between two regions (Additional file 1: Fig. S3).

A serial of SR-related ORs and AFs were evaluated based on two-day grouping of ODIs, and also separately delineated by sex, age and region over the whole study period (Fig. 3A‒D, Additional file 1: Tables S2‒S3). All-data based ODI risk functions showed a monotonic pattern, with the highest OR of 2.95 (95% CI 2.37‒3.66) at Day 10–11 and AF of 29.1% (95% CI 22.2%‒36.1%) for severe outcome of COVID-19 cases observed at Day 8–9 after adjusting for age, sex and region (Fig. 3A). A similar pattern was observed for the sex- or age-specific ORs and AFs of severe outcome, disclosing more robust adverse effects of ODI for younger cases (Fig. 3B‒3C). The effect of ODIs on SR was comparable between two regions until to prolonged ODI of Day 8‒9 (Fig. 3D). The magnitude of the association between ODI and severe disease was continuously elevated in Wuhan, with the highest OR and AF observed at Day 10–11 and Day 8–9, while the OR and AF were elevated to the highest level at Day 12–13 and Day 2–3 outside Wuhan. Lower ORs of ODI for severe outcome in cases aged ≥ 70 years were observed both in Wuhan and outside Wuhan (Additional file 1: Fig. S4).

There appeared to be only a slight decrease in CFR as ODI increased from 0 to 20 days (P > 0.05) (Fig. 2E). In addition to a slight increase in CFR in male cases, a slight decrease in CFR was also observed in all age and gender subgroups, with increasing ODI (Fig. 2F‒G). However, the linear correlation between CFR and ODI was significant both in Wuhan and outside Wuhan, with APCs of − 2.17 and − 12.12, respectively (Fig. 2H). When two regions were considered respectively, only female cases and cases aged 60‒69 years outside Wuhan had a significant positive association between CFR and ODI, with APCs of 6.99% and 7.88%, respectively (Additional file 1: Fig. S3).

A serial of CFR-related ORs and AFs were evaluated based on two-day grouping of ODIs, and were also separately delineated by sex, age and region over the whole study period (Fig. 3E‒H, Additional file 1: Table S4‒S5). Although ODIs of 2‒3 days were shown to have significant impacts on death outcome (P < 0.05), the magnitude of this risky effect was minor, with OR of 2.17 and AFs from 12.4%. Significant differences were observed for sex-, age-, and region-specific ORs and AFs of death outcome, disclosing the adverse effects of ODI on death outcome were only seen among male cases, younger cases aged 40‒59 years, and cases from Wuhan (Fig. 3F‒H). Notably, male cases displayed an increasing risk of COVID-19 death as the increase of ODI, with the highest OR of 1.39 (95% CI 1.05–1.85) and AF of 16.9% (95% CI 0.8–33.1%) at the ODIs of Day 2–3, while no such effect was observed for female cases (Fig. 3F, Additional file 1: Table S4‒S5). An increased risk of death from the increase of ODI was observed from patients aged 40–59 years, with the OR of 2.03 (95% CI 1.20–3.46) and AF of 33.6% (95% CI 9.0–58.2%) at the ODIs of Day 2–3, while no such effect was observed for other age groups (Fig. 3G, Additional file 1: Table S4‒S5). The effect of ODI on fatal outcome was only observed in Wuhan, with the OR of 1.31 (95% CI 1.03–1.67) and AF of 13.9% (95% CI − 0.3 to 28.2%) at the ODIs of Day 2–3, while no such effect was observed outside Wuhan (Fig. 3H, Additional file 1: Tables S4‒S5). Among the cases in Wuhan, higher ORs of ODI for fatal outcome were demonstrated in male than in female, in the 40–59 years group than in the ≥ 60 years group (Additional file 1: Fig. S5). Among the cases outside Wuhan, insignificant associations were observed in most two-day groups of ODI stratified by sex and age.

The estimated overall numbers of severe cases ranged from 1 236 (95% CI 683‒2 047) to 1 801 (95% CI 1 553‒2 076) under the 16 scenarios, which were significantly increasing with the prolonged ODI. If all cases were diagnosed within 10 days after symptom onset, the predicted severe cases were estimated to be 1.4 times of the scenario when all cases were diagnosed with 0 or 1 day after symptom onset (Fig. 4 and Additional file 1: Table S6). If all cases were diagnosed within 15 days after symptom onset, the predicted number of severe cases was further increased to 1 801 (95% CI 1 553‒2 076). For prediction of deaths in different scenarios, the estimated overall numbers ranged from 764 (95% CI 405‒1 422) to 844 (95% CI 692‒1 029), which slightly increased in relate to the prolonged ODI, and with much lower magnitude than those of severe cases. If all cases were diagnosed before 3 days after symptom onset, only little reduction of death cases was obtained, which was close to the actual number of death cases reported.