The influence of the COVID-19 pandemic on identifying HIV/AIDS cases in China: an interrupted time series study

Summary

Background

The COVID-19 pandemic has caused significant global public health challenges, and impacted HIV testing and reporting worldwide. We aimed to estimate the impact of COVID-19 polices on identifying HIV/AIDS cases in China from 2020 to 2022.

Methods

We used an interrupted time series (ITS) design and seasonal autoregressive integrated moving average intervention (SARIMA Intervention) model. Monthly reported data on HIV/AIDS cases were extracted from the National Bureau of Disease Control and Prevention of China from January 2004 to August 2022. Data on Stringency Index (SI) and Economic Support Index (ESI) from January 22, 2020 to August 31, 2022 were extracted from the Oxford COVID-19 Government Response Tracker (OxCGRT). Using these, a SARIMA-Intervention model was constructed to evaluate the association between COVID-19 polices and monthly reported HIV/AIDS case numbers from January 2004 to August 2022 using auto.arima () function from R. The absolute percentage errors (APEs) compared the expected numbers generated by the SARIMA-Intervention model with actual numbers of HIV/AIDS, and was the primary outcome of this study. A second counterfactual model estimated HIV/AIDS case numbers if COVID-19 hadn't occurred in December 2019, and the mean difference between actual and predicted numbers were calculated. All statistical analyses were performed in R software (version 4.2.1) and EmpowerStats 2.0 and a P < 0.05 was considered statistically significant.

Findings

The SARIMA-Intervention model indicated HIV/AIDS monthly reported cases were inversely and significantly correlated with stricter lockdown and COVID-19 related polices (Coefficient for SI = −231.24, 95% CI: −383.17, −79.32) but not with economic support polices (Coefficient for ESI = 124.27, 95% CI: −309.84, 558.38). APEs of the SARIMA-Intervention model for prediction of HIV/AIDS cases from January 2022 through August 2022, were −2.99, 5.08, −13.64, −34.04, −2.76, −1.52, −1.37 and −2.47 respectively, indicating good accuracy and underreporting of cases during COVID-19. The counterfactual model estimates between January 2020 and August 2022 an additional 1314 HIV/AIDS cases should have been established monthly if COVID-19 hadn't occurred.

Interpretation

The COVID-19 pandemic influenced the allocation and acquisition of medical resources which impacted accurate monthly reporting of HIV in China. Interventions that promote continuous HIV testing and ensure the adequate provision of HIV services including remote delivery of HIV testing services (HIV self-testing) and online sexual counseling services are necessary during pandemics in future.

Funding

Ministry of Science and Technology of the People's Republic of China (The grant number: 2020YFC0846300) and Fogarty International Center, National Institutes of Health, USA (The grant number: G11TW010941).

Research in context.

Evidence before this study

In the advent of COVID-19, China took a Dynamic COVID-zero approach to the pandemic which interrupted nation-wide HIV testing schedules. This will likely skew China's trajectory of achieving a global target of reporting 95% of all HV cases by 2030. Still, the exact impact on monthly reported HIV cases is unclear, therefore a systematic search of PubMed was conducted using search terms (COVID-19 OR SARS-CoV-2) AND (HIV OR AIDS) AND (policy OR measure OR strategy OR lockdown OR shelter OR restriction OR quarantine OR social distance OR economic) AND (China OR Chinese). A total of 296 results were retrieved between 1 January 2010 and 16 February 2023, and 14 studies were analyzed after title, abstract, and full-text screening. A previous ITS reported a decrease in the number of AIDS reported cases and deaths in the first month of lockdown (January 2020) in China. Other ITS studies indicated HIV/AIDS was negatively associated with the COVID-19 epidemic, but no correlation with government policies between January 2020 and December 2020 was established. A study analyzing aggregate Chinese HIV services data prior and in the advent of COVID-19 indicated interventions for COVID-19 diminished the volume of HIV testing and diagnoses. A retrospective cohort study among MSM indicated facility-based HIV testing and HIV self-testing (HIVST) sharply decreased after COVID-19. In contrast, an online survey of MSM across China implied overall HIV testing rates before and during COVID-19 measures were comparable, with HIVST making up for simultaneously decreased facility-based testing. Another online survey found HIV testing declined in February–April 2020 among MSM and never fully recovered since. Lastly, transaction data from a large online shopping platform in China revealed HIVST kit sales were significantly reduced while COVID-19 restrictions were in place, and restored when restrictions eased. Hence, the impact of the COVID-19 pandemic on identifying HIV/AIDS cases remains unclear and controversial.

Added value of this study

This interrupted time series study suggests COVID-19 related policies profoundly influenced HIV/AIDS testing, leading to reduced case reporting and higher rates of under surveilled HIV infections in China. To unveil such a finding, this study drew upon national data from the National Bureau of Disease Control and Prevention of China and the Oxford COVID-19 Government Response Tracker (OxCGRT) and forecasted long-term trends of HIV/AIDS case numbers if COVID-19 hadn't occurred in December 2019. These findings show stricter COVID-19 containment polices were indeed associated with lower HIV/AIDS case reporting, and an additional 1314 HIV/AIDS cases per month should have been expected if COVID-19 hadn't occurred.

Implications of all the available evidence

While China chased a national COVID-zero approach, the focus on identifying emergent HIV/AIDS cases was lost, leading to gross under-reporting of monthly HIV/AIDS numbers and underestimation of HIV incidence in China. This will likely lead to higher rates of HIV/AIDS infections in future. Balancing the distribution of medical resources between HIV and COVID-19 should be a greater focus of public health in China, and the provision of feasible HIV testing strategies like self-testing should be endorsed during lockdown periods.

Introduction

To optimize HIV prevention and treatment, the identification of people living with a new diagnosis of HIV is imperative for preventing transmission and achieving viral suppression.¹ According to the 2014 targets launched by the Joint United Nations Programme on HIV/AIDS (UNAIDS), it is a global goal that 95% of all PLWH individuals should be diagnosed by 2030,^2,3 although in 2022 this number is yet to be achieved in many parts of the world. In recent years, China expanded HIV screening and testing strategies, allowing HIV testing to increase over four-fold between 2009 and 2020 (55.6–240 million). Even so, delayed diagnosis of HIV infections from asymptomatic infected individuals is a great concern for HIV-control strategies,^4,5 and approximately 770,000 infected individuals were not virally unsuppressed in China, with 360,000 of these undiagnosed in 2018.⁶

Since its first reports from Wuhan in China in December 2019, Coronavirus disease 2019 (COVID-19) became a global pandemic, interrupting every aspect of normal life.⁷ Countries sought various means of virus control, but China approached with a “Dynamic COVID-zero” strategy, which in brief sought to find, control, and cure infectious persons within targeted regional cluster outbreaks.⁸ Whilst this strategy successfully contained domestic transmission of COVID-19 and avoided wide-spread outbreaks, the cost of such rigorous methods reverberates upon medical systems still today. This find indicates the cost and indirect population health and social impacts of this national response are needed to be further estimated.⁹ In particular, quarantine, social distancing, regional lockdowns of communities, and redirection of medical resources for COVID-19 prevented adequate HIV testing, surveillance of disease progression and medical personnel for the close monitoring and administration of anti-retroviral therapies (ART).⁷ As COVID-19 approaches its third year as a pandemic, defining its impact on the HIV/AIDS pandemic in China is of great importance.

The interrupted time series (ITS) studies became a popular study design during COVID for establishing causality between data points as it incorporates longitudinal data in scenarios where randomized controlled trails are not feasible.10, 11, 12 In addition, seasonal autoregressive integrated moving average intervention (SARIMA-Intervention) models emerged as a reliable method of forecasting that evaluated the effect of specific events, and was recently applied to evaluate the effect of HIV disease control interventions in South Africa.^10,13 Therefore, an ITS was conducted to evaluate the influence of COVID-19 policies on the diagnosis of new HIV/AIDS infections. To do this, monthly HIV/AIDS case numbers and continuous indices on COVID-19-related policy (the Oxford COVID-19 Government Response Tracker, OxCGRT) from public databases were collected to construct the SARIMA model. Comparing with a similar ITS,¹⁴ the current study estimated the impact of COVID-19 policies on HIV/AIDS identification precisely and continuously, to evaluate its effect on case numbers. The findings from this study suggest a definite impact of COVID-19 policies on HIV/AIDS underreporting and suggest stakeholders to balance the allocation of HIV resources and with future policies to minimize future impact.

Methods

Study design and data source

An ITS was conducted using monthly reported HIV/AIDS cases across 31 provinces in mainland of China using data available from the National Bureau of Disease Control and Prevention website (http://www.nhc.gov.cn/jkj/new_index.shtml).^15,16 This routine reporting system for notifiable infectious diseases was established by the Chinese government in the 1950s and was transformed from paper-based recording to web-based reporting in 2003.^15,16 In 1986 and 2020, HIV and COVID-19 respectively were identified as Class B notifiable diseases in China, meaning newly identified cases required reporting within 24 h.¹⁷ The Chinese government publishes the reported case numbers for all notifiable infectious diseases nationwide monthly on a publicly available website, excluding the sociodemographic characteristics of each case.¹⁶ This study extracted monthly cases of HIV/AIDS and COVID-19 from this database.

The intervention parameters for the SARIMA-Intervention model were Stringency Index (SI) and Economic Support Index (ESI), which are indicators of government measures toward COVID-19. These were extracted from OxCGRT, which provides cross-national and longitudinal data on government responses regarding to COVID-19 since January 2020.¹⁸ SI focuses on government policies related to restrictions on interpersonal physical contact as well as public information campaigns about COVID-19, and consists of 9 sub-indices: a) closings of schools and universities; b) closings of workplaces; c) cancelling public events; d) limits on gatherings; e) closing of public transport; f) orders to “shelter-in-place” and otherwise confine to the home; g) restrictions on internal movement between cities/regions; h) restrictions on international travel; and i) presence of public info campaigns.^19,20 Higher scores in SI indicate stricter enforcement of these policies. ESI quantified data about COVID-19 related economic compensation policies and included two sub-indices: a) record if the government is providing direct cash payments to people who lose their jobs or cannot work and b) record if the government is freezing financial obligations for households.^19,20 Higher scores in ESI indicate more economic support.

Considering this study is a secondary analysis that utilized de-identified data from publicly available databases, the study did not undergo ethics approval.

Procedure

Firstly, a Seasonal autoregressive integrated moving average intervention (SARIMA-Intervention) model was constructed based on monthly HIV/AIDS data and COVID-19 policy parameters ESI and SI to evaluated for the association between the COVID-19 policy parameters and monthly HIV/AIDS reported cases. We also utilized the SARIMA-Intervention model to predict the HIV/AIDS reported cases and estimate the accuracy of prediction by the absolute percentage errors (APEs). Secondly, a counterfactual model without the use of intervention parameters was built to further evaluate the effect of COVID-19 on identifying HIV/AIDS cases in the counterfactual scenario that COVID-19 didn't happen in December 2019. In ITS analysis, the counterfactual model forecasts the dependent variable in the absence of the specific interventions or other public health events (which is “counterfactual” scenario) and analyzes how the observed diverges from this forecast.¹⁰ The process of analysis was shown in Fig. 1.

Fig. 1.

Diagram of analysis procedure. a, the counterfactual refers to the hypothetical situation without the specific interventions or other public health events.

Statistical analysis

Pre-processing data

Since OxCGRT reported indices based on daily reports, monthly SI and ESI were calculated by averaging of daily SI and ESI scores per month from 22 January 2020. Monthly SI were transformed into four quartiles (Q1, Q2, Q3, and Q4) as ordinal categorical variables, and ESI was re-assigned to 0 (if ESI = 0), 1(if 0 < ESI ≤ 20), 2(if 20 < ESI ≤ 40) and 3 (if ESI > 40).²¹ Before January 2020, SI and ESI were classified as 0. From here, a time series of HIV/AIDS cases, SI and ESI were combined. We divided data into training set (January 2004–December 2021) and testing set (January 2022–August 2022) for SARIMA-Intervention model. A counterfactual model was designed where SARIMA-Intervention model neglected the effect of ESI and SI, where the training set included data from January 2004 to December 2019, and the testing set included data from January 2020 to August 2022.

Construction of model

Using auto.arima () function from forecast package in R,^22,23 the SARIMA-Intervention model was built based on the training set. The auto.arima () function in R was utilized to automatically choose an optimal set of parameters to build model, which compares the combinatorial spectrum of parameters according to the rule of the minimum Akaike's information criterion (AIC) or Bayesian information criterion (BIC).¹⁶ Only the optimal model is available as an output, hence no other models were able to be viewed and recorded in the results. In our study, we used Bayesian information criterion (BIC) in model selection, and Ljung–Box test was used to check whether the residual sequence was a white noise sequence (P > 0.05 if the residual sequence is white noise sequence).¹⁶ Furthermore, the mean absolute percentage error (MAPE) evaluated for the goodness-of-fit models, which is categorized as highly accurate forecasting (<10%), good forecasting (10%–20%), reasonable forecasting (20%–50%), and inaccurate forecasting (>50%).^16,24,25 In order to test the accuracy in prediction of model, we used absolute percentage error (APE), which was calculated by¹⁶:

$$\mathrm{A}\mathrm{P}\mathrm{E}=(\mathrm{A}\mathrm{c}\mathrm{t}\mathrm{u}\mathrm{a}\mathrm{l}\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}-\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r})/\mathrm{A}\mathrm{c}\mathrm{t}\mathrm{u}\mathrm{a}\mathrm{l}\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}$$

Accuracy was categorized as highly accurate forecasting (|APE%|<10), good forecasting (|APE| (%) = 10–20), reasonable forecasting (|APE| (%) = 20–50) or inaccurate forecasting (|APE| (%) >50). Furthermore, a counterfactual model was constructed based on the structure of SARIMA-Intervention model (SARIMA-Intervention model without the intervention part) and utilized to estimate the monthly reported number of HIV/AIDS cases during January 2020–August 2022 under the scenario that COVID-19 didn't occur in December 2019. The Ljung–Box test and MAPE were also calculated to test the residuals of the counterfactual model and evaluate for goodness-of-fit. We calculated the average additional HIV/AIDS cases per month in the counterfactual scenario to evaluate the potential loss on identifying HIV/AIDS cases during January 2020–August 2022. All statistical analyses were performed in R software (version 4.2.1) and EmpowerStats 2.0 (http://www.empowerstats.com). For more details, please check the R code analyzed in our study (Supplementary materials).

Role of the funding source

The funders played no roles in the study design, data collection, data analysis, data interpretation, writing of the report and the decision to submit the paper for publication.

Results

The visualization of the overall time-series data shows monthly HIV numbers fell in February 2020 (Fig. 2), around the same time that monthly reported cases of COVID-19 peaked (also in February of 2020, and April 2022 additionally, Fig. 2). Grossly, monthly HIV cases ceased to follow the same upward trend from April 2020 onwards that was evident ten years preceding (Fig. 2). Pearson correlation analysis showed monthly reported HIV/AIDS cases were inversely correlated with COVID-19 cases, and the number of HIV/AIDS climbed once COVID-19 cases were under control (r = −0.423, P = 0.016), indicating the COVID-19 pandemic influenced monthly reported HIV/AIDS cases. Monthly reported cases of HIV/AIDS and COVID-19 in China is quantitatively presented in Table 1 and Table 2, respectively.

Fig. 2.

Distribution of monthly reported HIV/AIDS and COVID-19 cases in China from January 2004 to August 2022.

Table 1.

Monthly reported newly diagnosed HIV/AIDS cases in China from January 2004 to August 2022.

Year	January	February	March	April	May	June	July	August	September	October	November	December
2004	60	170	153a	329	205	243	451	246	408	434	288	445
2005	287	212	346	411	594	897	706	637	954	536	517	521
2006	340	366	590	639	432	565	573	606	1269	441	545	608
2007	466	405	610	749	675	723	737	717	738	576	631	776
2008	658	686	1162	1161	1162	1077	1115	1156	1057	1077	964	1134
2009	926	948	1356	1394	1359	1551	1512	1474	1478	1718	2388	3209
2010	1663	1190	2411	2586	2619	3514	5930	3282	3620	2477	3322	3980
2011	1907	1294	3551	2806	3202	3817	5775	3912	3105	2867	4142	5434
2012	1232	2692	3879	3392	4281	5052	5552	4792	4727	4012	5078	5163
2013	2407	1838	4295	3806	3913	4122	4085	3878	4037	3284	4422	4404
2014	2245	2289	4221	3369	3823	4432	4354	3997	4611	3999	4464	5411
2015	2597	2294	4238	4299	4401	5089	4701	4268	4967	3955	5040	5707
2016	2862	2582	5255	4574	5041	5159	4459	4469	4877	4244	5743	6335
2017	2503	3325	5179	4140	5025	5916	4950	5400	5283	4485	6136	6622
2018	3309	2559	5331	4642	5593	5809	5289	5750	6155	5823	7622	7897
2019	3688	3587	6086	6277	6291	6642	6912	6404	6435	6207	7366	6735
2020	2759	2133	4808	5960	5484	6915	6124	5166	6927	4546	5824	6508
2021	3260	3051	5951	5283	5047	5978	3373	4710	5039	5357	6493	7490
2022	3109	3364	5020	3837	4490	5626	4667	4679	/	/	/	/
Average	1909.37	1841.32	3391.68	3139.68	3349.32	3848.79	3750.79	3449.63	3649.28	3113.22	3943.61	4354.39

^aData on March 2003 was extracted from Gazette of the Ministry of Health of the People's Republic of China.

Table 2.

Monthly reported COVID-19 cases in China from January 2020 to August 2022.

Year	January	February	March	April	May	June	July	August	September	October	November	December
2020	11,791	68,033	1730	995	143	517	803	721	356	583	545	529
2021	2493	348	305	454	451	670	1213	1893	1264	1081	1581	3490
2022	3825	3387	41,577	65,484	7547	1541	3919	13,855	/	/	/	/
Average	6036.33	23,922.67	14,537.33	22,311.00	2713.67	909.33	1978.33	5489.67	810.00	832.00	1063.00	2009.50

Intervention model performance and evaluation results

Information about the chosen SARIMA-Intervention and counterfactual models are detailed in Supplementary materials (Supplementary Table S1). The model chosen was Regression with SARIMA(1,1,1)(0,1,1)_[12] errors, with a Ljung–Box Q = 0.087 (P = 0.768), indicating the residuals sequence of the model was a white noise sequence and the model fully extracted information in the training set. This model forecasted the number of monthly HIV/AIDS cases from January to August of 2022. In this model, higher SI predicted a reduction in HIV/AIDS reporting per month (Coefficient = −231.244, 95% CI: −383.168, −79.319), although no statistically significant correlation was found between ESI and monthly HIV/AIDS reporting (Coefficient = 124.270, 95% CI: −309.838, 558.377). Table 3 and Fig. 3 display SARIMA-Intervention model monthly predictions. Absolute values of APE (%) within January, February, May, June, July, and August of 2022 are lower than 10%, indicating this model predicted precisely for this period, although absolute values of APE (%) in March and April of 2022 are higher than 10% (13.64% and 34.04% respectively), suggesting this was not a wholly accurate prediction. It was during this time monthly COVID-19 cases peaked (41,577 cases in March and 65,484 cases in April).

Table 3.

Forecast and actual number of HIV cases from January 2022 to August 2022 based on the SARIMA-Intervention model.

Date	Actual number	Predicted number (95% CI)	Absolute percentage error (APE, %)
Jan 2022	3109	3201.812 (2054.927, 4348.697)	−2.99
Feb 2022	3364	3193.048 (1919.693, 4466.404)	5.08
Mar 2022	5020	5704.601 (4376.877, 7032.324)	−13.64
Apr 2022	3837	5143.157 (3776.673, 6509.641)	−34.04
May 2022	4490	4613.732 (3213.106, 6014.358)	−2.76
Jun 2022	5626	5711.587 (4278.634, 7144.539)	−1.52
Jul 2022	4667	4730.839 (3266.564, 6195.114)	−1.37
Aug 2022	4679	4794.339 (3299.481, 6289.196)	−2.47

Fig. 3.

The actual number of reported HIV cases versus predicted HIV cases based on SARIMA-Intervention model. The shadings in the figure corresponds to 80% CI and 95% CI respectively.

Counterfactual model performance and evaluation results

Using monthly data of HIV/AIDS between January 2004 and December 2019, the counterfactual model SARIMA(1,1,1)(0,1,1)_[12] generated the expected number of HIV/AIDS cases per month during January 2020–August 2022 in the scenario that COVID-19 hadn't occurred. A Ljung–Box Q = 0.003 indicated the residuals sequence of the model was a white noise sequence (P = 0.958). As shown in Fig. 4, expected numbers of cases as predicted by the counterfactual model were lower than actual numbers. The independent sample t-test showed the mean of expected series predicted by the model was significantly higher than the actual series (t = −4.143, P < 0.001, Table 4) and the actual numbers in February, March, October and November of 2022 as well as July of 2021 are lower than the lower 95% confidence limits. The annual actual cumulative numbers for 2020, 2021 and 2022 (to the end of August) are 63,154, 61,032 and 34,792 respectively. And the annual expected cumulative numbers for 2020, 2021 and 2022 (to the end of August) are 73,720, 77,067 and 50,225 respectively. The total number of expected HIV/AIDS cases was 201,013 cases, and the actual number of cases reported during this time was 158,978. When the differences were taken, a missing 42,035 cases were identified in total, indicating an additional 1314 HIV cases should be expected in the monthly averages. This equals to approximately 26% of actual numbers in total between January 2020 and August 2022.

Fig. 4.

The actual number of reported HIV cases versus predicted HIV cases based on counterfactual model. The shadings in the figure corresponds to 80% CI and 95% CI respectively.

Table 4.

Prediction of HIV cases from January 2020 to August 2022 based on the counterfactual model.

Date	Actual number	Predicted number (95% CI)	P valuea
Jan 2020	2759	3707.683 (2664.65, 4750.716)	<0.001
Feb 2020	2133	3669.206 (2538.76, 4799.651)
Mar 2020	4808	6189.396 (5010.123, 7368.67)
Apr 2020	5960	6016.903 (4795.157, 7238.648)
May 2020	5484	6358.026 (5095.921, 7620.131)
Jun 2020	6915	6725.446 (5424.341, 8026.551)
Jul 2020	6124	6644.6 (5305.648, 7983.552)
Aug 2020	5166	6478.082 (5102.327, 7853.837)
Sep 2020	6927	6611.165 (5199.566, 8022.764)
Oct 2020	4546	6274.934 (4828.38, 7721.489)
Nov 2020	5824	7639.914 (6159.228, 9120.599)
Dec 2020	6508	7405.086 (5891.039, 8919.133)
Jan 2021	3260	4045.565 (2258.355, 5832.775)
Feb 2021	3051	3952.343 (2073.117, 5831.569)
Mar 2021	5951	6463.512 (4513.652, 8413.373)
Apr 2021	5283	6289.532 (4274.066, 8304.999)
May 2021	5047	6630.411 (4551.811, 8709.01)
Jun 2021	5978	6997.79 (4857.985, 9137.596)
Jul 2021	3373	6916.938 (4717.639, 9116.236)
Aug 2021	4710	6750.419 (4493.196, 9007.641)
Sep 2021	5039	6883.501 (4569.805, 9197.198)
Oct 2021	5357	6547.271 (4178.446, 8916.095)
Nov 2021	6493	7912.25 (5489.552, 10334.948)
Dec 2021	7490	7677.422 (5202.022, 10152.822)
Jan 2022	3109	4317.901 (1591.937, 7043.865)
Feb 2022	3364	4224.679 (1391.282, 7058.077)
Mar 2022	5020	6735.849 (3814.349, 9657.348)
Apr 2022	3837	6561.868 (3557.203, 9566.534)
May 2022	4490	6902.747 (3817.531, 9987.963)
Jun 2022	5626	7270.126 (4106.47, 10433.783)
Jul 2022	4667	7189.274 (3949.086, 10429.462)
Aug 2022	4679	7022.755 (3707.803, 10337.707)
Total	158,978	20,1012.595

^aThe independent sample t-test was conducted to compare the means of actual numbers and predicted numbers during January 2020–August 2022.

Discussion

This time series study evaluated the influence of Chinese COVID-19 government policies on monthly reported numbers of HIV/AIDS cases between January 2020 and August 2022 using two SARIMA models. It was uniquely identified that HIV/AIDS cases plummeted when stringent Chinese COVID-19 containment policies were in place, implying resources to seek and identify newly arising HIV/AIDS cases was insufficient and under reported. Alongside this, the absolute values of APE (%) in the SARIMA-Intervention model climbed above 10% in March and April of 2022, at the peak of monthly COVID-19 cases (41,577 and 65,484, respectively). The counterfactual model forecasts if COVID-19 hadn't occurred, an average additional 1314 HIV/AIDS cases should have been recorded monthly from January 2020 to August 2022. Our study identified potential underreporting of HIV/AIDS cases during COVID-19 pandemic in China, and numerically quantified the extent of loss in reporting. These findings suggest over a quarter of HIV/AIDS cases were never identified during the pandemic, suggesting a similar number of infectious individuals still dwell in Chinese communities without surveillance or acknowledgement of their disease state.

Despite the concerning finding that an estimated quarter of HIV/AIDS were not reported in 2022, exploration into this topic is complicated because COVID-19 containment measures directly restrict gatherings and access to public places. This could confound the propagation of normal HIV transmission trends, but also suspends resources where potentially infected individuals could seek HIV testing and ART.²⁶ Therefore, the probability that COVID-19 containment policies lowered HIV monthly numbers due to lower HIV testing and underreporting needs to be teased away from the probability transmission was lower due to social isolation. Findings from previous studies support the hypothesis for underreporting, with Shi et al.²⁷ finding from January to March 2020 HIV testing decreased by 49%, and of those positive just 71.4% were linked to services. Just 49.5% of positive cases received subsequent CD4 count testing, and while 91.6% of these received ART therapy, those that did not received CD4 count testing did not get ART. Overall, this lack of care to people living with HIV (PLWH) likely propagated infections in the community. Secondly, an online survey of 595 men who have sex with men (MSM) implied HIV testing was significantly lower in February–April 2020 (44.0%) than that November 2019–January 2020 (61.0%) when COVID-19 was not significantly influencing containment policies, supporting that HIV testing may be linked to lower access to services.²⁸ Lastly, the direct impact of COVID on delayed diagnosis of other infectious diseases were observed in China, with Wu et al.,²⁹ finding COVID-19 lockdowns led to poor and delayed access of tuberculosis (TB) infected individuals to diagnostic testing. Once COVID-19 numbers subsided, an uptick in TB notifications occurred. A similar phenomenon was observed in the current study, implying lower HIV/AIDS cases were not wholly the outcome of social isolation and absence of transmission. These findings together suggest HIV likely continued to propagate throughout the COVID-19 pandemic, and the findings of the current study probably reflect reality, although lower rates of transmission may have also been present.

The findings of this study highlight the reverberating effects COVID-19 has had on Chinese HIV services and this will likely influence the trajectory of China achieving the 2030 UNAIDS goal of reporting 95% of active HIV infections. Moreover, the public health implications of asymptomatic and sexually active PLWH can be disastrous if these continue uncontrolled and under-surveilled. Therefore, future efforts should ensure the sufficient provision of medical resources to maintain HIV care as COVID-19 outbreaks continue to occur. Considering the regional disparities in health-care resources availability and accessibility in China,^30,31 balancing the distribution of medical resources between COVID-19 and HIV should be a priority. For example, regions with relatively higher risk of HIV transmission like Yunnan, Guangxi, Sichuan and Guangdong should place heavy emphasis on continuity of HIV services alongside the containment of COVID-19 especially for those at higher risk of infection like serodiscordinant couples. In China non-governmental organizations (NGOs) care for marginalized and stigmatized populations which confer to higher HIV infection risks,¹⁷ therefore government organizations should openly cooperate with NGOs to ensure these services do not cease for HIV testing. Corresponding online sexual counseling services delivered by NGOs should be facilitated to ensure high-risk populations are followed.³²

Beyond government logistics, the risk of acquiring COVID-19 in health facilities has caused citizens to forgo HIV testing in countries such as Australia, China and Uganda, and presents as a social barrier to HIV testing services.33, 34, 35, 36 In the advent of such ambivalence, remote delivery of HIV testing services or self-testing (HIVST) presents itself as a convenient, home-based testing modality that is performed without the close contact of medical staff.^7,37 HIVST is a private alternative that makes HIV testing more palatable to those worried about stigma and discrimination within medical facilities,³⁸ although its incorrect use, misinterpretation, multiple testing kit options on the market and a lack of national guidance prevent it from being a simple substitution for facility-based testing.^17,39 Now more than ever, a Chinese government-led initiative for the coordination of HIVST is necessary to raise HIV testing rates and provide opportunities for people wishing to seek testing in a COVID-safe way. This simple intervention could solve many issues with resource distribution as the COVID-19 pandemic continues, and would allow for residual medical services to provide CD4 T cell count tests, ART therapies and ongoing remote sexual counselling to those that tested positive on home-based tests.^40,41

In this study, a SARIMA-Intervention model evaluated the impact of COVID-19 related policies on the identification of HIV/AIDS cases, and utilized a counterfactual model to explore the losses in case reporting. This compared pre-and post-pandemic HIV reporting trends, which allowed for a statistically significant effect of COVID-19 containment policies on case rates to be observed.¹³ The findings from this study contain unique knowledge about the impact of COVID-19 policies on HIV/AIDS in China and may be extrapolated to influence the future of governance for organizing HIV testing and medical resources in lockdowns, should they arise in future. With the stringent measures the Chinese government has persisted with, these findings will be of interest to many stakeholders globally.

Despite the impact of these findings, some limitations of this study should be acknowledged. Firstly, the prediction error of SARIMA model could increase when applied to longer prediction times and affect the accuracy of our results, which has led to the expansion of the 95% CI with increased prediction points in this study. In addition, the APEs of the SARIMA-Intervention model peaked in March and April in 2022 at the time of the COVID-19 case peak, indicating confounding factors could impact the accuracy of findings during this time. This could be due to frequency of risk behaviors during lockdown periods, but a lack of data on transmission and other influencing factors could influence the accuracy of models. Therefore, the accuracy of the model could be improved by introducing additional intervention series but the data for this was not available at the time of analysis. Despite these two concerns, the APE consisted mainly within the highly accurate range and so these findings likely represent the true impact of COVID-19 policies on HIV/AIDS reporting, and accurate HIV forecasting. In addition, this study focused on HIV/AIDS cases within the general Chinese population, however in higher risk areas like Guangdong, the influence of COVID-19 should be explored in a subgroup analysis. Furthermore, lockdown levels differed across China and HIV rates are also distributed differently across provinces. To guide tailored interventions for HIV-related medical services that are province-specific, there will need to be in-depth analyses for these provinces and geographic locations. Lastly, HIV/AIDS transmission is usually through sexual contact, and in the advent of heavy lockdown measures HIV transmission may have been very different to transmission trends in the scenario COVID-19 hadn't occurred. Therefore, while forecasting numbers in the counterfactual may provide insight to underreporting, it could potentially give false positive numbers. Despite this, throughout the pandemic serodiscordinant couples may have had decreased access to ART which puts them at higher risk of infection, and throughout the COVID-19 many people could disobey government policies unknowingly. Considering this study could not account for such confounders, these findings should be interpreted carefully and in the context of other emerging research.

In conclusion, the findings from this ITS study suggest COVID-19 containment polices in China had definite impacts on the detection of monthly HIV/AIDS cases, and these correlated with severity of lockdown measures. In future, the allocation of HIV medical resources in future COVID-19 lockdowns or pandemics should be prioritized to allow for the continual testing and surveillance of high-risk and infected populations. Such interventions will be imperative to the containment and control of HIV/AIDS transmission in China.

Contributors

TZ and ZW conceptualized the study. TZ oversaw data collection. TZ and ZC oversaw the curation of data. TZ and HL analyzed the data. TZ wrote the first draft of the manuscript. ZW, GB and ZJ helped in revising the manuscript. All authors in our study approved the final version of the manuscript for submission.

Data sharing statement

Relevant data related to monthly reported numbers of HIV/AIDS in China could be found on the webpage of the National Bureau of Disease Control and Prevention. And data of the Oxford Covid-19 Government Response Tracker could be found via the project GitHub repository¹⁸ (https://github.com/OxCGRT/covid-policy-tracker).

Declaration of interests

We declare no competing interests.