This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
Men who have sex with men (MSM) in China are disproportionately affected by the HIV epidemic, and medication adherence to antiretroviral treatment in this vulnerable population is suboptimal. To address this issue, we developed an app-based case management service with multiple components, informed by the Information Motivation Behavioral skills model.
We aimed to conduct a process evaluation for the implementation of an innovative app-based intervention guided by the Linnan and Steckler framework.
Process evaluation was performed alongside a randomized controlled trial in the largest HIV clinic in Guangzhou, China. Eligible participants were HIV-positive MSM aged ≥18 years planning to initiate treatment on the day of recruitment. The app-based intervention had 4 components: web-based communication with case managers, educational articles, supportive service information (eg, information on mental health care and rehabilitation service), and hospital visit reminders. Process evaluation indicators of the intervention include dose delivered, dose received, fidelity, and satisfaction. The behavioral outcome was adherence to antiretroviral treatment at month 1, and Information Motivation Behavioral skills model scores were the intermediate outcome. Logistic and linear regression was used to investigate the association between intervention uptake and outcomes, controlling for potential confounders.
A total of 344 MSM were recruited from March 19, 2019, to January 13, 2020, and 172 were randomized to the intervention group. At month 1 follow-up, there was no significant difference in the proportion of adherent participants between the intervention and control groups (66/144, 45.8% vs 57/134, 42.5%;
The intervention was well-received. Delivering educational resources of interest may enhance medication adherence. The uptake of the web-based communication component could serve as an indicator of real-life difficulties and could be used by case managers to identify potential inadequate adherence.
Clinicaltrial.gov NCT03860116; https://clinicaltrials.gov/ct2/show/NCT03860116
RR2-10.1186/s12889-020-8171-5
In China, men who have sex with men (MSM) are disproportionally affected by the HIV epidemic and remain one of the vulnerable populations susceptible to HIV infection via sexual transmission. The pooled prevalence of HIV among Chinese MSM from 2001 to 2018 was 5.7% (95% CI 5.4-6.1) [
The upsurge in mobile phone coverage and internet users in the past decade in China makes telemedicine and mobile health intervention highly feasible, which also applies to MSM, among whom the possession of cell phones and access to the internet is high [
Multicomponent app-based interventions are increasing in popularity [
Therefore, in this study, we aimed to conduct a process evaluation of our multicomponent app-based intervention from a quantitative perspective. We first examined the effect of the intervention by comparing medication adherence and IMB scores between the intervention and control groups and then described the intervention uptake of each component guided by the Linnan and Steckler [
Process evaluation was conducted alongside an ongoing open-label, parallel-group randomized controlled trial (RCT) in the HIV clinic in Guangzhou Eighth People’s Hospital in Guangzhou, China. Participants were recruited from March 19, 2019, to January 13, 2020. Eligible participants (1) were HIV-positive MSM aged ≥18 years, (2) had planned to initiate ART on the day of recruitment, and (3) had access to a personal smartphone and a private WeChat account to receive follow-up questionnaire links and the app-based intervention for the intervention group. More details regarding the recruitment procedure are described in a published protocol [
Eligible participants completed a baseline questionnaire on a tablet with assistance from field investigators and were then randomly assigned to the intervention or control group. The control group received a standard case management service, including a 20-minute education session before initiating ART and 4 hospital visits at months 1, 2, 3, and 6 for prescription refill, physical assessment, and meeting with case managers. Case managers would discuss the appointment schedule with each participant individually, and patients can reach case managers via phone calls during office hours to consult treatment-related concerns and questions or rearrange the appointment. The intervention group received the above standard service plus the app-based intervention, which consisted of the following 4 components, including web-based communication with case managers, educational articles delivery, supportive service information retrieval, and hospital visit reminders. The details of each intervention component are described in subsequent sections.
The design of the intervention was guided by the IMB model. It hypothesizes that individuals who are well-informed, motivated to act, and possess the necessary behavioral skills for effective action will be more likely to initiate and maintain healthy behaviors [
Web-based communication was achieved via a combination of an instant messaging app (WeChat) used by the patients and the Trusted Doctor app used by the case managers. The patients can initiate conversations or receive messages via the WeChat platform connected to the Trusted Doctor app. To avoid overburdening case managers, we set a limit that up to 5 messages could be initiated in 3 days by patients as default. Case managers were authorized to adjust the limit after the primary evaluation of the patient’s situation. One-on-one training on the use of the app was provided to all the case managers.
A total of 44 articles were sent in a prespecified order during the trial, of which 21 articles were sent during month 1. The delivered articles covered 12 themes and were categorized as primary and secondary according to the level of interest expressed by HIV-positive MSM on ART during the design phase. The primary category includes general introduction of the disease and treatment, tips about adherence, side effects of medication, and daily life adjustments. The secondary category includes psychological adaptation, disclosure to family or friends regarding one’s infection, and transmission prevention. Article titles, categories, and dates of delivery are detailed in Table S2 in
We embedded a set of supportive services on the WeChat platform, on which the patients could access the information by clicking on relevant tabs. The following 5 tabs were provided: guidance on government reimbursement, retrieval of a digital version of physical testing results, rehabilitation services, mental health services, and treatment seeking for regular diseases.
Four hospital visit reminders were scheduled at months 1, 2, 3, and 6 in the intervention group.
As suggested by the World Health Organization (WHO) guideline for ART, the first month of ART is especially important because immune reconstitution inflammatory syndrome and early adverse drug reactions can become potential barriers for patients to establish adequate adherence [
Adherence to ART medication was measured using a validated 3-item scale with high sensitivity to inadequate adherence [
Information, motivation, and behavioral skills were measured using the
We also collected information and motivation scores at baseline but did not collect behavioral skills, as it asked pill-taking scenarios (eg, “How hard or easy is it for you to take your HIV medications when your usual routine changes, for example, when you travel or when you go out with your friends?”), which does not apply to our participants who had not initiated the treatment at the time of the survey.
Process evaluation was guided by the Linnan and Steckler [
In this study, educational articles delivery, instructive message for supportive service information retrieval, and hospital visit reminders were automatically sent by the app, and their dose delivered was reported. Web-based communication and access to educational articles required interaction with participants, and the dose received was reported. Moreover, the web-based communication service was provided by case managers, and its fidelity was reported. Individual-level data on web-based communication and educational articles were provided by the Trusted Doctor company, whereas data on supportive service information can only be obtained via WeChat, in which only summative data are available. Therefore, web-based communication and educational articles were regarded as major intervention components and were primarily described and analyzed. A summary of the intervention component, corresponding process evaluation indicators, data sources, and type of data (ie, individual or summative) are listed in
Dose received was assessed based on the average number of messages received and sent by the participants. Trusted Doctor provided deidentified granular data for web-based communication: the content, time stamp, and identification number of the participants or case managers who sent or received the message. Table S5 in
As the assessment of fidelity can be flexible by definition, we developed a web-based patient-provider communication framework (Figure S2 in
For educational articles, the dose delivered and received were reported. During month 1, 21 selected articles were delivered to all participants in the intervention group. As data on whether the participants unsubscribed to the platform were not available, we presumed that all the articles were delivered as planned.
We used access to educational article links as a proxy for reading behavior. The dose received was measured by the number of access and access:delivery ratio (the number of access divided by the number of articles delivered) by article theme. Reading behaviors for primary and secondary categories were also constructed to assess the engagement of the participants. Reading behaviors were categorized as very high reading (ie, accessing every article delivered in month 1 at least once), high reading (ie, accessing only some of the articles with access:delivery ratio >1), adequate reading (ie, accessing only some of the articles with access:delivery ratio <1), and no reading engagement (not reading any of the articles in the specific category).
One reminder was automatically sent to all participants in the intervention group during the first month to inform them of the availability of the supportive service information embedded in the platform. Similar to the dose delivered by educational article delivery, dose delivered of supportive service information was presumed to be delivered as planned.
The dose received of supportive service information was measured by the number of clicks on the supportive service tab on the WeChat platform. As individual-level data were not available, only the overall number of clicks in each tab was reported.
The number of automatic reminder messages was reported as dose delivered, which was sent once in the first month.
At the month-1 survey, the intervention group was asked to rate how helpful the intervention was to their treatment and life on a 5-point scale, the options varying from “not helpful at all” to “very helpful.”
Contamination in the control group was measured by asking whether they engaged in web-based communication with research assistants (where the contact was established for delivering follow-up questionnaires) and whether they subscribed to other platforms for educational articles.
Descriptive statistics were generated for baseline variables, month-1 intermediate outcome (IMB scores), month-1 medication adherence, and indicators of process evaluation. Chi-square analysis and independent group 2-tailed
In addition to descriptive analysis, we conducted a series of exploratory analyses to test the month-1 efficacy of the intervention and to explore the potential mechanism between intervention uptake and month-1 behavioral outcomes. The month-1 efficacy of the intervention was examined in the full sample using logistic and linear regression for binary (ie, adherence) and continuous outcomes (ie, IMB scores), respectively, adjusting for unbalanced baseline variables between groups. The applicability of the IMB framework to our study sample was tested using structural equation modeling for the full sample. We also divided the control group by contamination and repeated the above analysis as a sensitivity analysis. Subsequently, we investigated the potential intervention mechanisms by examining the association between intervention uptake (ie, dose received from web-based communication and educational articles delivery) and medication adherence or IMB scores. First, we ran logistic or linear regression between baseline characteristics and outcomes in the control group to identify potential confounders. Then, we examined how intervention uptake was associated with the outcome of interest, adjusting for identified confounders. Missing data were imputed using multiple imputations by chained equations. Both text analysis and data analysis were performed using the R software (version 4.1.0; R Foundation for Statistical Computing).
The trial was approved by the Ethics Committee of the School of Public Health, Sun Yat-sen University (approval #003 in 2017). The RCT was registered at ClinicalTrials.gov under registration number NCT03860116. Written informed consent was obtained from all participants before the baseline assessment. For data analysis, all personal identifiers were removed and replaced by a research ID in password-protected data files to ensure confidentiality.
Between March 19, 2019, and January 13, 2020, we screened a total of 1690 patients who attended the HIV clinic for the first time, of which 344 were eligible and agreed to participate in our study (
Flowchart of the process evaluation study.
The baseline characteristics of the patients are presented in
At baseline, participants’ education level differed significantly between groups. Compared with the control group, the intervention group had a lower proportion of tertiary education (college degree or above, intervention vs control: 98/172, 57% vs 120/172, 69.7%;
Participants’ baseline characteristics (n=344).
Characteristics | All participants, n (%) | Control (n=172), n (%) | Intervention (n=172), n (%) | |||
|
.75 | |||||
|
18-24 | 86 (25.4) | 42 (25.1) | 44 (25.7) |
|
|
|
24-29 | 119 (35.2) | 62 (37.1) | 57 (33.3) |
|
|
|
30 | 133 (39.3) | 63 (37.7) | 70 (40.9) |
|
|
|
|
|||||
|
Middle school or lower | 55 (16) | 25 (14.5) | 30 (17.4) |
|
|
|
High school | 71 (20.6) | 27 (15.7) | 44 (25.6) |
|
|
|
College degree | 106 (30.8) | 63 (36.6) | 43 (25) |
|
|
|
Bachelor’s degree or above | 112 (32.6) | 57 (33.1) | 55 (32) |
|
|
|
.52 | |||||
|
Students | 43 (12.5) | 24 (14) | 19 (11) |
|
|
|
Public sector | 11 (3.2) | 6 (3.5) | 5 (2.9) |
|
|
|
Private sector | 157 (45.6) | 82 (47.7) | 75 (43.6) |
|
|
|
Other | 133 (38.7) | 60 (34.9) | 73 (42.4) |
|
|
|
.08 | |||||
|
<1000 | 61 (17.7) | 29 (16.9) | 32 (18.6) |
|
|
|
1000-5000 | 91 (26.5) | 37 (21.5) | 54 (31.4) |
|
|
|
5000-10,000 | 103 (29.9) | 61 (35.5) | 42 (24.4) |
|
|
|
>10,000 | 89 (25.9) | 45 (26.2) | 44 (25.6) |
|
|
|
.99 | |||||
|
Single | 199 (57.8) | 100 (58.1) | 99 (57.6) |
|
|
|
Not single | 145 (42.2) | 72 (41.9) | 73 (42.4) |
|
|
|
.73 | |||||
|
Homosexual or bisexual | 305 (88.7) | 151 (87.8) | 154 (89.5) |
|
|
|
Do not know or others | 39 (11.3) | 21 (12.2) | 18 (10.5) |
|
|
|
.10 | |||||
|
≥350 | 128 (46.0) | 69 (51.5) | 59 (41) |
|
|
|
<350 | 150 (54.0) | 65 (48.5) | 85 (59) |
|
|
|
.99 | |||||
|
Free first-line regimen | 287 (83.4) | 144 (83.7) | 143 (83.1) |
|
|
|
Other regimens | 57 (16.6) | 28 (16.3) | 29 (16.9) |
|
aIndicates significant between-group difference.
At the month-1 follow-up, 45.8% (66/144) of the participants in the intervention group and 42.5% (57/134) in the control group reported adequate adherence without statistical significance (
Effects of intervention on intermediate and health outcomes of the 2 groups.
Outcomes | ORa or β coefficient (95% CI) | |||
|
||||
|
Adherence | 1.31 (0.80 to 2.13) | .28 | |
|
||||
|
Information score | −0.19 (−1.30 to 0.92) | .73 | |
|
Motivation score |
|
|
|
|
Behavioral skills | 0.93 (−1.23 to 3.10) | .40 |
aOR: odds ratio.
bIndicates OR.
cIndicates β coefficient.
dIndicates significant effect size.
In total, 120 (70%) of 172 participants were engaged in web-based conversations with case managers, generating 1664 messages. On average, each participant who engaged in web-based conversation sent 7 messages and received 6 during month 1. After coding these messages into dialogues, one-third (37/120, 30.8%) had only 1 dialogue with case managers, and two-fifths (46/120, 38.3%) had 2 or 3 dialogues, with the rest having ≥3 dialogues.
A total of 401 dialogues were coded to assess the fidelity of web-based communication, including content (ie, practical or emotional), completeness (whether the questions were solved), timeliness (defined as whether the initial message was replied within 24 hours), and style (ie, formal language use and emoticon use;
Characteristics of web-based patient-provider communication. (A) The content of the web-based conversation. (B) The completeness and timeliness of the web-based conversation reply. (C) The style of the web-based conversation (unit: dialogue).
Of the 401 dialogues, 374 (93.2%) received a text-based response and the rest responded with web links. Of the 374 dialogues, 366 (97.9%) had practical content focusing on problem solving, and the primary concern in web-based conversations was the side effects of medication (114/374, 30.5%). Other topics included changes in daily life habits because of treatment (87/374, 23.3%), hospital visit arrangement (66/374, 19.4%), the influence of imperfect medication-taking behavior (36/374, 9.4%), necessary physical examinations (32/374, 8.6%), other potential treatment regimens (13/374, 3.5%), transmission prevention (10/374, 2.8%), comorbidity (4/374, 1.3%), and other undefined topics (12/374, 3.2%). A small portion (35/374, 9.4%) of the dialogues involved emotional content, that is, patients expressed negative feelings or case managers tried to cheer up patients. Regarding completeness, 94.9% (355/374) of the issues were solved directly, whereas the rest were not solved or referred to other professionals. In terms of timeliness, the majority (326/374, 87.2%) replied within 24 hours. Regarding the style of web-based conversation, formal language was used in 39.6% (148/374) of the dialogues (eg, greetings before questions, patients’ expressions of gratitude, or case managers’ expression of willingness to help). Emoticon was less common, as it was used in only 12% (45/374) of the dialogues.
Of the 172 participants in the intervention group, 158 (91.9%) accessed at least 1 delivered article. The total number of accesses was 3128. Articles from the primary category were more popular than those from the secondary category (
Dose received of the educational articles by theme. (A) The number of access for education articles of each theme. (B) Access:delivery ratio for articles of each theme. ART: antiretroviral treatment.
The services tab named
A reminder was sent to each participant in the intervention group. In total, 172 reminder messages were sent.
Out of 144 participants in the intervention group who completed the web-based survey, 47 (32.6%) rated the intervention platform to be “very helpful” and 77 (53.5%) rated as “helpful.”
Using structural equation modeling, we solidified the application of the IMB models to the full sample (Figure S4 in
We also examined the relationship between intervention uptake (ie, web-based communication and educational articles) and IMB scores in the intervention group, as illustrated in
Hypothesis and tested associations about intervention mechanism. (A) Directed acyclic graph of the intervention mechanism in full sample. (B) Directed acyclic graph of the intervention components and outcomes in the intervention group. Each directed line represents a hypothesized causal relationship. The solid black line represents significant association tested. The dashed line represents nonsignificance.
Association between intervention uptake and month-1 adherence in the intervention group. Logistic regression models adjusting for factors associated with adherence in the control group (ie, education and monthly income) were constructed to obtain odds ratio (OR) and 95% CI.
Variables | OR (95% CI) | ||||||||
|
|||||||||
|
|
||||||||
|
|
Practical support | 0.98 (0.86-1.12) | .81 | |||||
|
|
Emotional support | 0.79 (0.54-1.16) | .23 | |||||
|
|
||||||||
|
|
Problem solved | 0.97 (0.84-1.12) | .70 | |||||
|
|
||||||||
|
|
Replied within 24 hours | 0.97 (0.83-1.13) | .72 | |||||
|
|
||||||||
|
|
Either side using emoticons | 0.83 (0.55-1.25) | .37 | |||||
|
|
Either side using formal language | 1.01 (0.83-1.23) | .90 | |||||
|
|||||||||
|
|
|
|
||||||
|
|
|
|
|
|||||
|
|
|
General introduction |
|
|
||||
|
|
|
Medication tips |
|
|
||||
|
|
|
Side effects |
|
|
||||
|
|
|
Daily life | 1.22 (0.92-1.61) | .17 | ||||
|
|
|
|||||||
|
|
|
Inadequate reading | Reference | —b | ||||
|
|
|
Adequate reading | 0.97 (0.41-2.28) | .94 | ||||
|
|
|
Very high reading | 0.43 (0.18-1.06) | .07 |
aOdds ratios and
bNot applicable as inadequate reading was the reference level.
Association between intervention uptake and scores of month-1 Information Motivation Behavioral skills model constructs in the intervention group. Baseline scores were adjusted in the regression model. For information scores, whether engaged in web-based communication was adjusted.
Variables | Information, β (95% CI) | Motivation, β (95% CI) | Behavioral skills, β (95% CI) | ||||||
|
|||||||||
|
|
||||||||
|
|
Practical support | –0.1 (–0.33 to 0.13) | – |
–0.32 (–0.75 to 0.11) | ||||
|
|
Emotional support | –0.37 (–0.79 to 0.05) | – |
– |
||||
|
|
||||||||
|
|
Problem solved | –0.13 (–0.34 to 0.09) | – |
–0.30 (–0.76 to 0.16) | ||||
|
|
||||||||
|
|
Replied within 24 hours | –0.14 (–0.39 to 0.1) | – |
–0.35 (–0.79 to 0.1) | ||||
|
|
||||||||
|
|
Either side using emoticons | 0.02 (–0.47 to 0.51) | – |
–0.67 (–2.2 to 0.86) | ||||
|
|
Either side using formal language | –0.03 (–0.34 to 0.27) | – |
–0.15 (–0.77 to 0.47) | ||||
|
|||||||||
|
|
0.02 (–0.11 to 0.14) | –0.07 (–0.23 to 0.09) | 0.05 (–0.21 to 0.31) | |||||
|
|
|
0.03 (–0.14 to 0.19) | –0.11 (–0.32 to 0.09) | 0.09 (–0.26 to 0.43) | ||||
|
|
|
General introduction | –0.15 (–0.76 to 0.47) | –0.35 (–1.15 to 0.45) | 0.27 (–0.97 to 1.52) | |||
|
|
|
Medication tips | –0.05 (–0.61 to 0.52) | –0.2 (–0.9 to 0.49) | 0.16 (–0.98 to 1.31) | |||
|
|
|
Side effects | 0.24 (–0.3 to 0.78) | –0.37 (–1.04 to 0.31) | 0.62 (–0.45 to 1.68) | |||
|
|
|
Daily life | 0.05 (–0.44 to 0.55) | –0.29 (–0.92 to 0.34) | –0.33 (–1.35 to 0.69) | |||
|
|
|
|||||||
|
|
|
Inadequate reading | Reference | Reference | Reference | |||
|
|
|
Adequate reading | –0.75 (–2.61 to 1.1) | –0.34 (–2.69 to 2.01) | 0.80 (–2.71 to 4.31) | |||
|
|
|
Very high reading | –0.35 (–2.25 to 1.55) | 1.62 (–0.94 to 4.18) | 0.48 (–3.35 to 4.31) |
aβ coefficients and 95% CI that indicated significance was italicized.
During the study period, 29.7% (51/172) of the participants in the control group initiated web-based communication with our research team via WeChat contact, and 51% (69/134) of participants in the control group reported subscribing to other WeChat platforms for educational materials.
In this study, we conducted a process evaluation of an app-based intervention for HIV-positive MSM. We found no statistically significant difference in medication adherence between the intervention and control groups at the month-1 follow-up. The intervention components were well received, especially web-based communication and educational articles delivery. The uptake of educational articles was associated with better adherence. Guided by the IMB model, we found that the app-based intervention helped maintain the motivation score among the participants, whereas the motivation score showed a decreasing trend in the control group. However, web-based communication was associated with lower motivation scores in the intervention group.
Detecting null results of medication adherence between the intervention and control groups is not uncommon in app-based interventional studies, in which studies investigate the efficacy of an app in addition to usual care, as in this study. In a small (N=50) RCT conducted among HIV-positive people with substance use disorder, DeFulio et al [
Compared with previous studies, we hypothesized that one of the reasons for our null result may be the so-called
Key intervention components were well-received, with 69.8% (120/172) participants engaging in web-based communication and 91.9% (158/172) participants accessing educational articles. Most of the web-based conversations revolved around problem-oriented content, which echoes the most frequently visited themes of educational articles, such as medicine taking, side effects of ART medication, and hospital visits. Together with the decreases in information score in both groups (Figure S3 in
Under the IMB framework, we examined the potential mechanisms of our key intervention components, and several unexpected findings are worth our attention. First, the hypothesis that the association between reading engagement and adherence is through the information-behavioral skills-adherence pathway (Figure S4 in
It is encouraging that the majority (124/144, 86.1%) of participants in the intervention group rated the intervention as helpful or very helpful to their life in the month-1 survey. Nevertheless, the well-received web-based conversation cannot be achieved without the dedication of case managers. The timeliness of web-based conversation is an advantage of our app-based intervention compared with other prevalent contact channels, such as email. During our training for case managers, the timeliness of web-based conversation was emphasized, as previous studies showed that timely responses from health care workers for web-based inquiries were highly valued by patients [
This study had several limitations. First, process evaluation can only be conducted at month-1 follow-up, which is short and may lead to our null findings. This was because of the discontinued collaboration with the technology company. Therefore, month-1 adherence, which can be used for investigating mechanisms, was used as a proxy for short-term efficacy measures. However, according to ART guidelines, the first month after ART initiation is critical for adequate adherence to medicine-taking behavior [
The strength of this study was the objective measures for intervention uptake rather than self-reported measures, which could yield more reliable evidence regarding the uptake of web-based conversations and reading of educational articles. In particular, web-based communication data in this study were collected and analyzed in a quantitative manner, guided by a self-designed framework. Previous studies on web-based communication usually included self-reported web-based communication behaviors or beliefs, and the findings were therefore subject to recall or measurement bias. The increasing popularity of web-based medical consultation in China [
The intervention was well-received. Delivering educational resources may enhance medication adherence. The uptake of the web-based communication component could serve as an indicator of real-life difficulties and could be used by case managers to identify potential inadequate adherence.
The theoretical base and conceptual model of intervention design, details of intervention delivery for each component, supplementary methodology regarding data processing, and supplementary results.
antiretroviral treatment
information, motivation, and behavioral skills
Life Windows Information–Motivation–Behavioral Skills Antiretroviral Therapy Adherence Questionnaire
men who have sex with men
odds ratio
randomized controlled trial
The Joint United Nations Programme on HIV and AIDS
World Health Organization
This work was funded by the National Natural Science Foundation of China (grant 71774178), Science and Technology Planning Project of Guangdong (grant 2017A020212006), and Guangzhou Science and Technology Project (grant 201607010368). The authors would like to thank all the participants and case managers that took an active part in this study.
The data sets used and analyzed during this study are available from the corresponding author upon reasonable request.
XF and KN contributed equally to this study. XF and KN drafted and revised the manuscript. XF, CL, and HZ collected the data. XF analyzed the data. JG, JTFL, CH, JL, LL, and YH conceptualized the study. All authors have read and approved the final manuscript.
None declared.