HIV/AIDS late presentation and its associated factors in China from 2010 to 2020: a systematic review and meta-analysis

Background Late presentation to HIV/AIDS care presents serious health concerns, like increased transmission and high healthcare costs, increased mortality, early development of opportunistic infection, increased risk of antiretroviral therapy drug resistance. Despite the effort to contain the HIV/AIDS epidemic, LP has remained an impediment to individual immune reconstitution and public health. Objective This review aimed to estimate the prevalence and determine the factors associated with late presentation to HIV/AIDS care. Methods We searched PubMed, Web of Science, China National Knowledge Infrastructure (CNKI), Chinese Wanfang, and Weipu database for articles published from 2010 to 2020. We utilized I2 statistics and Q-test to estimate heterogeneity between studies. Random-effects meta-analysis models were used to calculate the aggregate odds ratio of late presentation to HIV/AIDS care. Results Of 9563 titles and abstracts retrieved, 189 were identified as potentially eligible and 39 fulfilled the inclusion criteria. The pooled prevalence of late presentation to HIV/AIDS care was 43.26%. The major risk factors were patients ≥ 50 years old (OR = 2.19, 95% CI: 1.85–2.58; I2 = 97.44%), married (OR = 1.50, 95% CI: 1.35–1.68; I2 = 96.58%), with heterosexual contact as risk factor for infection (OR = 1.91, 95% CI: 1.73–2.11; I2 = 90.74%) and diagnosed in medical institutions (OR = 2.35,95% CI: 2.11–2.62; I2 = 96.05%). In middle or low HIV prevalence areas, patients ≥ 50 years old (P = 0.01), married (P < 0.01) and diagnosed in medical institutions (P = 0.01) were more likely to be presented late than in high prevalence areas. From 2016–2020, the OR of patients who were married and diagnosed in medical facilities were significantly lower than before (P < 0.01). Conclusion Patients ≥ 50 years old, married, with heterosexual contact as risk factor for infection, and diagnosed in medical institutions were risk factors of LP. Gender had no significant relationship with LP. In middle or low prevalence areas, patients who were ≥ 50 years old, married, and diagnosed in medical institutions were more likely to be presented late than in other areas. Married patients and those diagnosed in medical institutions after 2015 have a lower risk of LP than before. Supplementary Information The online version contains supplementary material available at 10.1186/s12981-021-00415-2.


Introduction
Late presentation (hereon in referred to as LP) to HIV care remains a challenge to HIV prevention and treatment in the world. In Guatemala from 2000 to 2015 [1], 81.1% of new diagnoses were considered late presentations. The prevalence of LP is estimated to range between 36.9% in Estonia to 64.2% in Poland in Europe during 2010-2016 [2]. In China, the percentage of patients with late HIV presentation ranged from 35.5 to 42.1% from 2010 to 2014 [3]. LP may lead to some grave consequences both for individuals and the society, such as increased mortality, development of opportunistic infection [4], increased risk of antiretroviral therapy (ART) drug resistance [5], high healthcare costs [6], and increased transmission because of unawareness of infection status [7]. Some comprehensive strategies, such as extensive study, free testing, and prompt treatment initiation have been taken [8,9]. However, only 75.7% of people who live with HIV know their infectious status by the end of 2019 in China.
Improving early presentation is of great importance for AIDS prevention and care. Early presentation means early entry to HIV care. Study evidence showed that patients shtarting early ART could have a near-normal life expectancy, provided that they start treatment before their CD4 count decreases below 200 cells/µl [10]. Early ART can also limit the HIV reservoir size [11,12]. And the smaller the HIV reservoir at treatment interruption, the better the post-treatment control [13]. To some degree, ART also can prevent sexual HIV transmission in both homosexual and heterosexual individuals [14].
To our knowledge, there is only two published systematic review and meta-analysis on HIV late presentation and its predictors [15,16]. In China, there are no such publications. Therefore, it is necessary to comprehend the situation of LP to HIV care and propose an effective program to promote early presentation in China. This meta-analysis aimed to estimate the pooled prevalence of LP and determine the risk factors from 2010 to 2020 in China. We hope to provide evidence for comprehensive prevention and testing strategy.

Literature search strategy
We searched PubMed, Web of Science, China National Knowledge Infrastructure (CNKI), Chinese Wanfang, and Weipu databases for articles published in English and Chinese from 2010 to 2020. Using the following Boolean term to search the databases: (TS = "HIV/ AIDS" OR TS = "human immunodeficiency virus" OR TS = "Acquired immunodeficiency syndrome") AND (TS = "late presentation" OR TS = "late diagnosis" OR TS = "late testing" OR TS = "delay presentation" OR TS = "delay diagnosis" OR TS = "delay testing") in Chinese and (TS = "HIV/AIDS") AND (TS = "late entry" OR TS = "Advanced HIV disease" OR TS = "late presentation" OR TS = "late Diagnosed") in English. We also retrieved the studies referenced in all included studies to obtain further related studies. Our analyses followed the 2009 Preferred Reporting Items for Systematic reviews and Meta-Analyses(PRISMA) statement.

Inclusion and exclusion criteria
Articles included met all the following criteria: (1) observational studies, including cross-sectional, case-control, and cohort studies; (2) studies reported the prevalence and associated factors of LP to HIV care in China from 2010 to 2020; (3) sufficient data present to estimate the odds ratios (ORs) with 95% confidence intervals (CIs); (4) studies defined late presentation of HIV/AIDS according to the European definition or other definitions recognized appropriate. Articles were excluded based on at least one of the following: (1) studies that do not meet the criteria above; (2) studies absence of original data, such as the number of patient with different factors; (3) studies published based on the same or overlapping data; (4) review, case report or meeting report.

Data extraction and quality assessment
The studies searched were managed by Endnote (version X9) and de-duplicated. The studies were first screened by title and abstracts independently by two investigators according to the inclusion and exclusion criteria. If there were any conflicts between the results, all three authors of this review would screen the full text to discuss and resolve it before reaching a consensus. Moreover, the references cited in all included studies were screened by two investigators independently. Information retrieved from each eligible study comprised of:-title, first author, publication year, study design, study period, study time, the definition of LP, geographical locations, source population, number of subjects in each category, the value of odds ratios(OR) and it's 95% CI. The Newcastle-Ottawa quality-assessment scale (NOS) was employed to assess the quality of those included studies. We evaluated the quality of evidence using the "grading of recommendations assessment, development, and evaluation" (GRADE) approach.

Statistical analysis
We estimated heterogeneity between studies with I 2 statistics and Q-test. Random-effect models were administered in testing significant heterogeneity (P < 0.10 or I 2 ≥ 50% implied statistically significant heterogeneity). Otherwise, fixed-effect models were applied. Subgroup analysis focused on study region and period. The study regions were categorized into high prevalence regions (including Yunnan, Guangxi, Henan, Sichuan, Xinjiang, and Guangdong provinces) and others according to the proportion of HIV infected patients in the nationally reported cases [17]. χ 2 test was employed to estimate the proportion difference between the subgroups. Sensitivity analysis was employed to assess the sensitivity of each included study. Publication bias was measured using the Harbord test and funnel plots (P > 0.05 represented no publication bias). We conducted all statistical analysis with Stata 16.0 and SPSS 24.0.

Study review and selection
A total of 4381 Chinese articles and 5182 English articles were obtained (CNKI 2624; Wanfang 1373; Weipu 384; PubMed 3931; Web of Science 1251). After the removal of duplicates, 6462(67.6%) remained. A total of 189 articles were eligible following titles and abstracts screening. After the full-text screening, 39 papers were finally included (36 Chinese and 3 English studies). The flowchart of studies identified by the search is in Fig. 1.

General characteristic of the included studies
All of the 39 studies were case-control studies. According to the NOS assessment, the scores of all studies were ≥ 5, which denoted good quality. Included papers were from 2010 to 2020. The study regions covered 21 provinces or municipalities: Guangxi, Guangdong, Sichuan, Xinjiang, Yunnan, Jiangsu, Taiwan, Anhui, Beijing, Fujian, Gansu, Guizhou, Jiangxi, Liaoning, Shandong, Shaanxi, Shanxi, Tianjin, Zhejiang, Chongqing, and Hubei. We classified all study regions into high HIV prevalence or middle/ low HIV prevalence groups based on HIV prevalence. Among these provinces and municipalities, Guangxi, Guangdong, Sichuan, Xinjiang, Yunnan are high HIV  Table 1. were more likely to be diagnosed late. While male was not a risk factor for LP (OR = 1.02, 95% CI: 0.90-1.15; I 2 = 95.45%). In high prevalence areas, patients ≥ 50 years old (P = 0.01), married (P < 0.01) and diagnosed in medical institutions (P = 0.01) were less likely to be presented late than in middle or low areas.

The quality of evidence
The quality grade of age, infection routes, and sample sources were high. The grade score of gender and marital status was moderate. An additional file shows this in more detail (see Additional file 1).

Sensitivity analysis
A leave-one-out sensitivity analysis was adopted to examine the possible cause of heterogeneity across the studies involved in the analysis. The sensitivity analysis results suggested that none of the individual studies influenced the initial total results.

Publication bias
The conventional funnel plots indicated showed almost no publication bias in the meta-analysis. An additional file shows this in more detail (see Additional file 2). We used Harbord-test to confirm the result and found no statistically significant differences (P > 0.05) ( Table 4).

Discussion
We conducted 39 publications to identify the related factors of LP in China. It showed that the overall LP proportion from 2010 to 2020 in China was 43.26%. Patients ≥ 50 years old, married, with heterosexual contact as risk factor for infection, and diagnosed in medical institutions were risk factors of LP. Gender had no statistically significant relationship with LP. In high prevalence areas, patients who were ≥ 50 years old, married, and diagnosed in medical institutions were less likely to be presented late than in other areas. It suggested the need for targeted measures to reduce the occurrence of LP in different regions. Additionally, we have made some suggestions on prevention and policy making of LP to HIV care based on these data.
In the general health environment, the elderly and female should be associated with reduced odds of LP because they have better health seeking behaviours. However, in our study the results were different. In China, partly of the HIV positive female did not realize the risk because they were infected by their husband. That might lead to the result after pooling the articles. Studies have shown that the elderly have limited access and ability to obtain and understand HIV/AIDS prevention information [57][58][59]. Additionally, the elderly tend to ignore HIV infection due to various comorbidity symptoms    Fig. 2 The forest plot of the association between age and late presentation. The midpoint and length of each segment indicated the OR and 95% confidence interval. The diamond shape revealed the pooled OR clinics and pre-surgery. Hence the current work of consulting and testing still needs improvement. For this part of patients, we believe that enriching the consulting and testing methods can effectively improve the poor situation, such as carrying out online consulting and starting high-sensitivity self-testing.         Fig. 6 The forest plot of the association between sample sources and late presentation. The midpoint and length of each segment indicated the OR and 95% confidence interval. The diamond shape revealed the pooled OR In subgroup analysis, the proportion of LP in high epidemic areas is higher than in middle or low regions. When comparing the occurrence of LP in two time periods in middle or low epidemic areas, the latter period was significantly higher than the former. Action plan for AIDS containment and prevention in China in 12th Five-Year put forward measures to expand the coverage of publicity and education, comprehensive intervention, testing and treatment [62]. Even though great efforts had done to expand the coverage of testing and treatment of HIV, LP is still a pressing problem in high prevalence regions. In recent years, the situation of LP in middle or low prevalence areas had become more severe than before. Thus, we should continue to expand the coverage of testing and treatment. The government should invest more funds in high prevalence regions to conduct focused testing or even census for high-risk groups to find more patients as early as possible. Besides, we recommend the exemption   There are some other findings in this study. Firstly, previous studies on the influencing factors of LP to HIV care mainly focused on social demographic determinants. There are still many other related factors to be explored, such as behavioral factors, AIDS knowledge level, access to testing and ART, policy, and social support. Secondly, there are few studies on the ART and immune reconstitution of LP patients. There are several limitations. The criteria for LP for the included publications were different, and the results may deviate from the real world. Therefore, there is an urgent need for a consensus definition of LP to facilitate full use of the actual material to reflect the problems and find solutions. Secondly, we included four articles in the 2016-2020 group, which may have led to bias in our results.
Generally, LP remains an obstacle to the prevention and treatment of HIV/AIDS in China. Targeted public health interventions to improve early entry into HIV care are urgently needed. We still have a lot to do for HIVrelated policy-making, testing strategy, and health education in the future.

Conclusion
Patients ≥ 50 years old, married, with heterosexual contact as risk factor for infection, and diagnosed in medical institutions were risk factors of LP. Gender had no significant relationship with LP. Although the country have expanded the coverage of testing and treatment of HIV through great efforts, LP is still a pressing problem in high prevalence regions. In middle or low prevalence areas, patients who were ≥ 50 years old, married, and diagnosed in medical institutions were more likely to be presented late than in other areas. Patients married and diagnosed in medical institutions after 2015 have a lower risk of LP than before. Governments should also take measures to expand the coverage of education, testing, and treatment of HIV.