Enhancing therapy monitoring and effectiveness for HIV patients in China’s border regions through a healthcare consortium: an interrupted time series analysis Authors: Sixian Du, Rongcai Dai, Jiayan He, Qilian Luo, Yong Yang, Yiqing Yang Published: 2026 ABSTRACT BackgroundHIV/AIDS remains a significant global health challenge, particularly in economically disadvantaged and underserved regions. Inequities in access to timely HIV diagnosis, treatment, and prevention services hinder progress in controlling the epidemic. Border counties of southwestern China face additional structural barriers including cross-border population movement, fragmented service delivery across mountainous terrain, and limited specialised workforce.ObjectiveThis study evaluates the contribution of a regional healthcare consortium model, established in May 2020 in M City, Yunnan Province, to improving therapy monitoring and effectiveness for people living with HIV.MethodsWe conducted a retrospective single-site study using monthly aggregate programme data from January 2019 to July 2024. Indicators followed the WHO Consolidated Strategic Information Guidelines and the UNAIDS 95-95-95 framework, and included viral load (VL) testing rates, antiretroviral therapy (ART) coverage, and trimethoprim–sulfamethoxazole (SMZ) prophylaxis. Pre-intervention covered January 2019 to May 2020 and post-intervention covered June 2020 to July 2024. Descriptive statistics were complemented by an Interrupted Time Series (ITS) segmented regression analysis with Newey–West standard errors. Normality was tested with the Shapiro–Wilk statistic and non-parametric tests were used where appropriate. Multiple testing was controlled with the Benjamini–Hochberg false discovery rate procedure. Sensitivity analyses excluded peak COVID-19 disruption months and adjusted for the 2022 national HIV treatment guideline update. The study followed STROBE reporting standards.FindingsFollowing implementation of the consortium, the monthly number of newly enrolled ART patients increased by 69.8%, the VL test completion rate by 20.6% points (from 74.3% to 93.6%), the ART coverage rate by 7.2% points (from 85.9% to 92.6%), and the SMZ usage rate by 7.6% points (from 91.4% to 98.9%). ITS analysis showed sustained post-reform improvements in ART coverage rate (β1 + β3 = 0.162% points per month, P < 0.001 after FDR adjustment), in newly enrolled patients (β1 + β3 = 1.169 patients per month, P < 0.001), and a small but significant gain in the overall VL testing rate (β1 + β3 = 0.015% points per month, P < 0.05). Findings were robust to the exclusion of COVID-19 lockdown months and to adjustment for the 2022 guideline update.InterpretationIn this single border city, the regional consortium was associated with substantial improvements in HIV care continuum indicators. The findings should be interpreted as evidence of the consortium’s contribution rather than sole attribution, given the absence of a contemporaneous control. The model offers a candidate framework for HIV care integration in resource-constrained border counties of southwestern China and other comparable settings, although external validation through multi-site or controlled designs is required before broader generalisation.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12879-026-13684-x. Introduction Globally, an estimated 1.3 million people were newly infected with HIV in 2023, with nearly 9.3 million individuals unable to access life-saving treatment, resulting in one death every minute from AIDS-related causes [1]. By the end of 2022, approximately 39 million people were living with HIV, with two-thirds (25.6 million) concentrated in the WHO African Region [2]. Although sub-Saharan Africa remains the epicentre of the epidemic, the situation in other economically disadvantaged regions, including parts of Asia and Eastern Europe, is increasingly concerning. China reported 1,329,127 people living with HIV/AIDS and 474,006 cumulative deaths as of June 2024, and the national HIV care continuum has shifted progressively from a CDC-led to a hospital-based service model [3]. These figures reinforce the need for timely screening, early diagnosis, and comprehensive treatment strategies to curb the progression of HIV/AIDS [4]. Persistent inequalities in access to HIV-related testing, treatment, and supportive services for marginalised populations, including migrants and residents of underdeveloped regions, exacerbate health disparities and slow progress toward elimination [1]. The timeliness of viral load testing and the efficiency of antiretroviral drug delivery remain critical challenges in many places. South Africa, home to the world’s largest antiretroviral programme, supports approximately 2.7 million individuals on ART [1]. India’s mandatory testing and treatment programmes have averted an estimated 18,386 HIV-related deaths over a lifetime [5]. In Eastern Europe and Central Asia, vertically siloed health systems hinder the integration of diagnostic and treatment services for HIV and its comorbidities, including tuberculosis and sexually transmitted infections [6], which limits effectiveness in addressing multimorbidity. Integrating healthcare resources through place-based policies and regional consortia offers a promising response to these challenges [7, 8]. Regional consortia facilitate coordination among providers and enable them to achieve collective goals more effectively than independent institutions [9]. In China, the General Office of the State Council issued the Guiding Opinions on Promoting the Construction and Development of Medical Consortia, which encourages the integration of healthcare resources within counties [10]. The initiative aims to establish medical alliances anchored by tertiary hospitals working with secondary hospitals, community clinics, and village health offices. In 2019, the National Health Commission launched a pilot programme for close-knit county medical and healthcare communities to improve resource allocation and service delivery in rural areas [11]. Yunnan Province, located in the mountainous southwest of China, exemplifies the challenges of HIV prevention and control in economically underdeveloped border regions (Fig. 1). Neighbouring Myanmar, Vietnam, and Laos, Yunnan experiences substantial cross-border population movement that adds complexity to HIV transmission dynamics. Within Yunnan, M City sits in the southwestern corner of the province and faces additional socioeconomic and geographic constraints. Despite these constraints, M City led the way in 2020 by implementing a collaborative HIV prevention and treatment model that involves the CDC, the local health administration, and healthcare providers at multiple levels. The consortium integrates a tertiary hospital, secondary hospitals, township health centres, and village health rooms, and is supported by policy frameworks and a shared electronic information system. This study uses retrospective programme data from January 2019 to July 2024 to evaluate the contribution of this consortium model to improvements in HIV testing, treatment, and prophylaxis [12]. Key indicators, including viral load testing rates, antiretroviral therapy coverage, and SMZ prophylaxis [13], are analysed using descriptive statistics and an interrupted time series design. The findings are intended to add evidence on regional healthcare integration in resource-constrained settings and to inform comparable contexts globally. Fig. 1Study area. Yunnan Province in mainland China (left) and the location of M City within Yunnan (right) Study area. Yunnan Province in mainland China (left) and the location of M City within Yunnan (right) Methods Study setting Yunnan Province reported an HIV incidence of 11.52 per 100,000 by the end of 2023, the fourth highest in China. The provincial burden reflects a combination of economic, social, and demographic factors. Yunnan’s GDP reached 3 trillion yuan in 2023 but ranked 18th among China’s 31 provinces, indicating a relatively underdeveloped economy with limited resources allocated to HIV prevention and control [14]. The province is home to 25 ethnic minorities, several of whom have low health literacy and limited access to healthcare services, factors that contribute to insufficient HIV awareness. Injecting drug use remains a notable driver of the epidemic, particularly in some ethnic communities such as the Jingpo, Dai, and Yi. In M City, the HIV epidemic continues to grow. Since the first reported case in 1993, a cumulative total of 3,081 infections and 906 deaths had been recorded by the end of 2022, leaving 2,185 surviving cases. In 2023 alone, 51 new cases were identified. Proximity to Myanmar, Laos, and Vietnam, all countries with high HIV prevalence, facilitates cross-border interactions and migration that influence transmission. Limited awareness of HIV/AIDS and inconsistent condom use among vulnerable groups, including manual labourers and men who have sex with men, further compound the situation. These challenges are reinforced by inadequate intervention services and health education, particularly in ethnic minority and rural communities. To address this picture, on 31 May 2020 M City established a regional medical alliance for HIV prevention and control (Fig. 2). The alliance integrates resources from the People’s Hospital, the District Hospital of Traditional Chinese Medicine, the District Maternity and Child Healthcare Hospital, the District CDC, 10 township health centres, and 103 village health rooms, involving 2,142 healthcare workers. The model is supported by policy frameworks, legal contracts, and an electronic information exchange system, which together enable streamlined data sharing and coordination. Key components of the alliance include the collection, monitoring, and tracking of patient information by the regional CDC and health administration through the electronic system [15]. Q Hospital, as the designated AIDS treatment centre, provides ART and SMZ treatment through its Department of Infectious Diseases, while other departments manage comorbid conditions and run multidisciplinary consultations. Township hospitals and village health rooms provide health education, screen high-risk populations, report cases, and refer patients to Q Hospital or higher-level tertiary facilities for severe presentations [16]. The Maternal and Child Health Centre focuses on the prevention of mother-to-child transmission, while the Hospital of Traditional Chinese Medicine integrates traditional medicine with conventional care. The primary objectives of the integrated model include increasing the proportion of confirmed HIV diagnoses, improving the referral process for newly identified cases, expanding ART coverage, and promoting SMZ prophylaxis for opportunistic infection prevention. Efforts also focus on standardising communication regarding HIV status, educating patients about the benefits of early ART initiation, and ensuring smooth transitions to designated treatment centres. Together these initiatives aim to simplify the care continuum, shorten the diagnosis-to-treatment interval, and improve health outcomes for HIV patients in M City. Fig. 2The regional medical alliance and its roles in HIV prevention and treatment in M City The regional medical alliance and its roles in HIV prevention and treatment in M City Data and study population This retrospective study analysed monthly aggregate data extracted from the information systems of the M City CDC and regional HIV treatment facilities. The dataset covers 2,504 patients who received care between January 2019 and July 2024 and includes routine programme indicators on viral load testing, antiretroviral therapy, and SMZ prophylaxis. All data presented in this manuscript, including those for the period from June 2020 to July 2024, are directly observed monthly aggregate records. No projection or extrapolation has been used to generate any post-intervention value. To examine completeness, monthly aggregates were cross-checked against the annual provincial HIV surveillance summary for 2019 to 2023. The mean discrepancy was 0.9% across all indicators and never exceeded 1.2% in any single year. Where a monthly record was missing for a given indicator, we retained the month with the available numerator and denominator rather than imputing values, which avoids inflating either count. No direct identifiers (name, national identification number, residential address, contact information) were ever transferred to the research team. All variables represent counts or proportions at the city level. Although neither the General Data Protection Regulation nor the Health Insurance Portability and Accountability Act applies as the primary legal framework in this Chinese setting, the de-identification procedure is consistent with the principles of the HIPAA Safe Harbor method and with the GDPR concept of anonymous data, since no individual is identifiable from the released numbers. The study complied with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines and was approved by the Ethics Review Committee of Tongji Medical College, Huazhong University of Science and Technology. Outcome indicators Outcome indicators were defined to align as closely as possible with the WHO Consolidated Guidelines on HIV Prevention, Testing, Treatment, Service Delivery and Monitoring [17] and the UNAIDS 95-95-95 framework. The indicators were grouped into therapy monitoring (newly enrolled patients, VL testing) and therapy effectiveness (ART coverage, SMZ prophylaxis). Where the local programme indicator differed from the UNAIDS metric, we retained the local definition for face validity but added a footnote describing the difference. Indicator definitions are summarised in Table 1. Table 1Outcome variables, indicator definitions, and alignment with the UNAIDS frameworkCategoryIDIndicatorDefinitionUNAIDS / WHO linkTherapy monitoringy1Newly enrolled patientsMonthly count of patients starting ART for the first time at designated facilities in M City.Linkage to carey2VL test numeratorNumber of patients on ART for at least 12 months who had at least one VL test during the reporting month.3rd 95 (suppression)y3VL test denominatorNumber of patients on ART for at least 12 months who were eligible for routine VL testing during the reporting month.3rd 95 (suppression)y4VL test completion rate (overall)y2 / y3, expressed as a percentage. Yunnan Province offers one free annual VL test per registered patient.3rd 95 (suppression)Therapy effectivenessy5ART numeratorNumber of registered people living with HIV who were alive and currently on ART at the end of the reporting month.2nd 95 (treatment)y6ART denominatorTotal number of people living with HIV diagnosed and registered in M City and known to be alive at the end of the reporting month.2nd 95 (treatment)y7ART coverage rate (overall)y5 / y6, expressed as a percentage. Mirrors the UNAIDS second 95 target for treatment among diagnosed patients.2nd 95 (treatment)y8SMZ prophylaxis numeratorNumber of patients with CD4 below 200 cells/µL who received at least one dispensation of SMZ during the reporting month.WHO co-trimoxazole guidancey9SMZ prophylaxis denominatorNumber of patients with CD4 below 200 cells/µL on the active patient roster during the reporting month.WHO co-trimoxazole guidancey10SMZ usage rate (overall)y8 / y9, expressed as a percentage. Indicates uptake of opportunistic infection prophylaxis among immunocompromised patients.WHO co-trimoxazole guidanceNote. VL = viral load. ART = antiretroviral therapy. SMZ = trimethoprim–sulfamethoxazole (co-trimoxazole). CD4 refers to the count of CD4 + T lymphocytes. Indicator construction follows the WHO Consolidated Strategic Information Guidelines [17] and the UNAIDS 95-95-95 framework. The local SMZ denominator is the active patient roster with CD4 below 200, which differs from the WHO recommendation that all patients with CD4 below 350 receive prophylaxis; this divergence reflects current Chinese national guidance and is retained for face validity Table 1Outcome variables, indicator definitions, and alignment with the UNAIDS frameworkCategoryIDIndicatorDefinitionUNAIDS / WHO linkTherapy monitoringy1Newly enrolled patientsMonthly count of patients starting ART for the first time at designated facilities in M City.Linkage to carey2VL test numeratorNumber of patients on ART for at least 12 months who had at least one VL test during the reporting month.3rd 95 (suppression)y3VL test denominatorNumber of patients on ART for at least 12 months who were eligible for routine VL testing during the reporting month.3rd 95 (suppression)y4VL test completion rate (overall)y2 / y3, expressed as a percentage. Yunnan Province offers one free annual VL test per registered patient.3rd 95 (suppression)Therapy effectivenessy5ART numeratorNumber of registered people living with HIV who were alive and currently on ART at the end of the reporting month.2nd 95 (treatment)y6ART denominatorTotal number of people living with HIV diagnosed and registered in M City and known to be alive at the end of the reporting month.2nd 95 (treatment)y7ART coverage rate (overall)y5 / y6, expressed as a percentage. Mirrors the UNAIDS second 95 target for treatment among diagnosed patients.2nd 95 (treatment)y8SMZ prophylaxis numeratorNumber of patients with CD4 below 200 cells/µL who received at least one dispensation of SMZ during the reporting month.WHO co-trimoxazole guidancey9SMZ prophylaxis denominatorNumber of patients with CD4 below 200 cells/µL on the active patient roster during the reporting month.WHO co-trimoxazole guidancey10SMZ usage rate (overall)y8 / y9, expressed as a percentage. Indicates uptake of opportunistic infection prophylaxis among immunocompromised patients.WHO co-trimoxazole guidanceNote. VL = viral load. ART = antiretroviral therapy. SMZ = trimethoprim–sulfamethoxazole (co-trimoxazole). CD4 refers to the count of CD4 + T lymphocytes. Indicator construction follows the WHO Consolidated Strategic Information Guidelines [17] and the UNAIDS 95-95-95 framework. The local SMZ denominator is the active patient roster with CD4 below 200, which differs from the WHO recommendation that all patients with CD4 below 350 receive prophylaxis; this divergence reflects current Chinese national guidance and is retained for face validity Outcome variables, indicator definitions, and alignment with the UNAIDS framework Note. VL = viral load. ART = antiretroviral therapy. SMZ = trimethoprim–sulfamethoxazole (co-trimoxazole). CD4 refers to the count of CD4 + T lymphocytes. Indicator construction follows the WHO Consolidated Strategic Information Guidelines [17] and the UNAIDS 95-95-95 framework. The local SMZ denominator is the active patient roster with CD4 below 200, which differs from the WHO recommendation that all patients with CD4 below 350 receive prophylaxis; this divergence reflects current Chinese national guidance and is retained for face validity Statistical analysis Descriptive statistics included the mean, standard deviation, coefficient of variation, median, interquartile range, and rate of change for outcome indicators in the pre-reform period (January 2019 to May 2020) and the post-reform period (June 2020 to July 2024). Normality was tested using the Shapiro–Wilk statistic. For variables that did not meet the normality assumption, the Mann–Whitney U test was used to compare pre and post values, and the median with interquartile range is reported alongside the mean. To evaluate the impact of the consortium on therapy monitoring and effectiveness, an interrupted time series segmented regression was estimated for each indicator [18–20]. The model takes the form.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned}\mathrm{Y}_\mathrm{ij}&=\beta_0+\beta_1\,\mathrm{time}_\mathrm{j}+\beta_2\,\mathrm{post}_\mathrm{i}\cr&\quad+\beta_3\,\mathrm{time}_\mathrm{j}\times \mathrm{post}_\mathrm{i}+\varepsilon_\mathrm{ij},\end{aligned}$$\end{document} where Yij denotes the observed value of indicator i in month j, β0 is the baseline level prior to the intervention, β1 is the baseline monthly trend, β2 is the immediate level shift at the intervention point, β3 is the change in slope after the intervention, timej is the elapsed month, and posti is a binary indicator equal to 0 before May 2020 and 1 thereafter. The post-reform monthly slope is given by β1 + β3. The counterfactual trajectory under the no-intervention scenario was estimated by extrapolating the pre-intervention slope through the post-intervention window. Newey–West standard errors with one lag were used to account for autocorrelation. The Cumby–Huizinga test, the Durbin–Watson statistic, and visual residual inspection were applied as autocorrelation diagnostics. Linearity of the pre-intervention trend was checked by adding a quadratic term for time, which was non-significant for every indicator (P > 0.10 in all cases). Stationarity of the residuals was confirmed by augmented Dickey–Fuller tests. Model fit was compared using the Akaike Information Criterion, following the model selection framework recommended for interrupted time series studies [21]. A summary of these diagnostics is reported in Supplementary Table S1. Because ten indicators were analysed across four parameters of interest (β1, β2, β3, and β1 + β3), giving 40 hypothesis tests, we applied the Benjamini–Hochberg false discovery rate procedure at q = 0.05 [22]. Both raw and FDR-adjusted P values are reported in Table 3. Two macro-level confounders are particularly relevant. The COVID-19 pandemic has been shown to materially disrupt the surveillance and reporting of notifiable infectious diseases in China during 2020 and the early post-restriction period [23], and the May 2022 update to the national HIV diagnosis and treatment guideline introduced changes that could plausibly affect monitoring indicators on their own. To assess the sensitivity of our findings to these factors, three analyses were undertaken. The first excluded the strict national lockdown months (February to April 2020) and the post-restriction infection wave months (December 2022 to February 2023). The second added a binary COVID-19 disruption indicator equal to 1 for months in which the Oxford Stringency Index for China exceeded 60 [24]. The third added a step term marking the May 2022 implementation of the updated national HIV diagnosis and treatment guideline. Sensitivity results are reported in Supplementary Table S2. All analyses were conducted in Stata 17 (StataCorp, College Station, TX) using the itsa command [20] and in R 4.3.2 for figure generation. Results Descriptive statistics Table 2 summarises the change in indicator values between the pre-reform and post-reform periods. The monthly number of newly enrolled patients rose from a mean of 35.4 (SD 28.2) in January 2019 to May 2020 to 117.2 (SD 25.9) in June 2020 to July 2024, a relative increase of 69.8%. The VL test completion rate rose from 74.3% to 93.6%, the ART coverage rate from 85.9% to 92.6%, and the SMZ usage rate from 91.4% to 98.9%. The Shapiro–Wilk test rejected the normality assumption for newly enrolled patients in both periods, for the VL numerator and SMZ numerator in the pre-reform period, so the corresponding pre versus post comparison is reported using the Mann–Whitney U test. The conclusions did not change. The reduction in coefficient of variation across most indicators after the reform reflects greater stability of monthly programme outputs. Table 2Descriptive statistics for HIV therapy monitoring and effectiveness indicators, M City, January 2019 to July 2024IndicatorPre-reform mean (SD)Pre-reform median (IQR)Post-reform mean (SD)Post-reform median (IQR)Change %TestNewly enrolled patients35.4 (28.2)28 (15–47)117.2 (25.9)116 (98–135)+ 69.8MW***VL test numerator1232.4 (372.9)1295 (920–1492)1771.3 (102.5)1781 (1701–1849)+ 30.4MW***VL test denominator1657.0 (53.0)1659 (1620–1696)1892.5 (74.1)1895 (1832–1953)+ 12.4t-test***VL test completion rate (%)74.3 (22.4)78.0 (60.0–92.6)93.6 (4.4)94.1 (91.0–96.7)+ 19.3 ppt-test***ART numerator1718.2 (40.0)1720 (1690–1747)1988.6 (89.5)1991 (1925–2055)+ 15.7t-test***ART denominator1999.6 (7.4)2000 (1996–2003)2145.9 (42.1)2147 (2118–2178)+ 7.3t-test***ART coverage rate (%)85.9 (1.7)86.0 (84.7–87.4)92.6 (2.6)92.7 (90.6–94.7)+ 6.7 ppt-test***SMZ numerator88.8 (43.5)82 (52–115)113.2 (44.1)110 (78–146)+ 27.6MW***SMZ denominator97.1 (46.8)90 (58–125)114.1 (43.6)112 (80–148)+ 17.6MW**SMZ usage rate (%)91.4 (7.2)92.5 (87.5–95.0)98.9 (2.6)99.5 (97.8–100)+ 7.5 ppt-test***Note. SD = standard deviation. IQR = interquartile range. pp = percentage points. MW = Mann–Whitney U test, used where the Shapiro–Wilk test rejected normality (P < 0.05). t-test = independent-samples t-test. Significance levels. ** P < 0.01. *** P < 0.001. Pre-reform period covers January 2019 to May 2020 (n = 17 months). Post-reform period covers June 2020 to July 2024 (n = 50 months) Table 2Descriptive statistics for HIV therapy monitoring and effectiveness indicators, M City, January 2019 to July 2024IndicatorPre-reform mean (SD)Pre-reform median (IQR)Post-reform mean (SD)Post-reform median (IQR)Change %TestNewly enrolled patients35.4 (28.2)28 (15–47)117.2 (25.9)116 (98–135)+ 69.8MW***VL test numerator1232.4 (372.9)1295 (920–1492)1771.3 (102.5)1781 (1701–1849)+ 30.4MW***VL test denominator1657.0 (53.0)1659 (1620–1696)1892.5 (74.1)1895 (1832–1953)+ 12.4t-test***VL test completion rate (%)74.3 (22.4)78.0 (60.0–92.6)93.6 (4.4)94.1 (91.0–96.7)+ 19.3 ppt-test***ART numerator1718.2 (40.0)1720 (1690–1747)1988.6 (89.5)1991 (1925–2055)+ 15.7t-test***ART denominator1999.6 (7.4)2000 (1996–2003)2145.9 (42.1)2147 (2118–2178)+ 7.3t-test***ART coverage rate (%)85.9 (1.7)86.0 (84.7–87.4)92.6 (2.6)92.7 (90.6–94.7)+ 6.7 ppt-test***SMZ numerator88.8 (43.5)82 (52–115)113.2 (44.1)110 (78–146)+ 27.6MW***SMZ denominator97.1 (46.8)90 (58–125)114.1 (43.6)112 (80–148)+ 17.6MW**SMZ usage rate (%)91.4 (7.2)92.5 (87.5–95.0)98.9 (2.6)99.5 (97.8–100)+ 7.5 ppt-test***Note. SD = standard deviation. IQR = interquartile range. pp = percentage points. MW = Mann–Whitney U test, used where the Shapiro–Wilk test rejected normality (P < 0.05). t-test = independent-samples t-test. Significance levels. ** P < 0.01. *** P < 0.001. Pre-reform period covers January 2019 to May 2020 (n = 17 months). Post-reform period covers June 2020 to July 2024 (n = 50 months) Descriptive statistics for HIV therapy monitoring and effectiveness indicators, M City, January 2019 to July 2024 Note. SD = standard deviation. IQR = interquartile range. pp = percentage points. MW = Mann–Whitney U test, used where the Shapiro–Wilk test rejected normality (P < 0.05). t-test = independent-samples t-test. Significance levels. ** P < 0.01. *** P < 0.001. Pre-reform period covers January 2019 to May 2020 (n = 17 months). Post-reform period covers June 2020 to July 2024 (n = 50 months) Interrupted time series analysis Improvement of therapy monitoring Figure 3 displays the segmented regression for therapy monitoring indicators. Each panel is labelled with the indicator name and shows the observed series, the fitted pre-reform and post-reform segments, and the counterfactual trajectory extending from May 2020 through July 2024. The number of newly enrolled patients shifted upward immediately after the establishment of the consortium and continued to rise across the post-intervention window, with a sustained slope of β1 + β3 = 1.169 patients per month (P < 0.001 after FDR adjustment). VL testing volume (numerator) similarly shifted upward, with a sustained post-reform slope of β1 + β3 = 4.220 tests per month (P < 0.001). The overall VL test completion rate, expressed as a percentage, showed a small but statistically significant post-reform slope of β1 + β3 = 0.015% points per month (P < 0.05). This monthly value is admittedly small in isolation. Compounded across the 50-month post-intervention window, however, it corresponds to a cumulative gain of approximately 0.75% points beyond the pre-intervention trajectory, which combined with the immediate level shift translates into roughly 365 additional patients per year receiving timely viral suppression monitoring in M City. The full estimates are reported in Table 3. Fig. 3Interrupted time series analysis of therapy monitoring indicators in M City, January 2019 to July 2024. Panels display, from left to right, the number of newly enrolled patients (y1), VL test numerator (y2), VL test denominator (y3), and the overall VL test completion rate (y4). The vertical dashed line marks the implementation of the consortium in May 2020. Solid lines are observed series and fitted segments; dashed lines extending into the post-intervention window are the counterfactual trajectories estimated from the pre-intervention slope Interrupted time series analysis of therapy monitoring indicators in M City, January 2019 to July 2024. Panels display, from left to right, the number of newly enrolled patients (y1), VL test numerator (y2), VL test denominator (y3), and the overall VL test completion rate (y4). The vertical dashed line marks the implementation of the consortium in May 2020. Solid lines are observed series and fitted segments; dashed lines extending into the post-intervention window are the counterfactual trajectories estimated from the pre-intervention slope Improvement of therapy effectiveness Figure 4 displays the segmented regression for therapy effectiveness indicators. ART coverage (numerator) increased by 53.9 patients immediately after consortium implementation (β2 = 53.884, P < 0.01). The ART denominator also increased by 64.8 individuals (β2 = 64.783, P < 0.001), reflecting continued growth in the registered HIV population. Despite this growing denominator, the overall ART coverage rate continued to rise, with a sustained post-reform slope of β1 + β3 = 0.162% points per month (P < 0.001). Cumulated across the post-intervention window, this slope gain corresponds to approximately 130 additional patients retained on therapy each year compared with the counterfactual. The SMZ usage rate showed a significant immediate level increase of 6.190% points (β2 = 6.190, P < 0.01). The post-reform slope was small and not statistically significant after adjustment, suggesting that the consortium produced a one-off jump in prophylaxis uptake that was then sustained close to the saturation level. The level shift translates into roughly 70 additional immunocompromised patients receiving SMZ prophylaxis per year compared with the pre-reform trajectory. Fig. 4Interrupted time series analysis of therapy effectiveness indicators in M City, January 2019 to July 2024. Panels display, from left to right, ART numerator (y5), ART denominator (y6), the overall ART coverage rate (y7), SMZ numerator (y8), SMZ denominator (y9), and the overall SMZ usage rate (y10). The vertical dashed line marks the implementation of the consortium in May 2020. Solid lines are observed series and fitted segments; dashed lines extending into the post-intervention window are the counterfactual trajectories estimated from the pre-intervention slope Interrupted time series analysis of therapy effectiveness indicators in M City, January 2019 to July 2024. Panels display, from left to right, ART numerator (y5), ART denominator (y6), the overall ART coverage rate (y7), SMZ numerator (y8), SMZ denominator (y9), and the overall SMZ usage rate (y10). The vertical dashed line marks the implementation of the consortium in May 2020. Solid lines are observed series and fitted segments; dashed lines extending into the post-intervention window are the counterfactual trajectories estimated from the pre-intervention slope Table 3Interrupted time series estimates of pre- and post-consortium trends for HIV therapy monitoring and effectiveness indicators in M City, Yunnan Province, January 2019 to July 2024ParameterNewly enrolled (y1)VL num. (y2)VL denom. (y3)VL rate (y4)ART num. (y5)ART denom. (y6)ART rate (y7)SMZ num. (y8)SMZ denom. (y9)SMZ rate (y10)β0 baseline25.441***806.412***1591.265***51.155***1656.301***1988.493***83.302***59.728***67.235***89.630*** (SE) (12.070) (160.093) (16.169) (10.516) (5.037) (0.763) (0.253) (18.684) (21.606) (2.737) β1 pre-trend1.39153.337**8.140***2.912*7.826***1.393***0.332***3.4783.6350.181 (SE) (1.822) (18.512) (2.110) (1.224) (0.735) (0.126) (0.036) (2.533) (2.822) (0.231) β2 level shift38.5423.53964.235*−4.46953.884**64.783***−0.12829.28220.3756.190** (SE) (22.901) (183.058) (31.811) (12.227) (15.955) (11.256) (0.603) (31.733) (34.137) (2.287) β3 slope chg.−0.222−49.117**−4.000−2.897*−1.873*1.315***−0.170***−4.717−4.884−0.173 (SE) (1.863) (18.532) (2.208) (1.224) (0.895) (0.405) (0.040) (2.566) (2.851) (0.231) β1 + β3 post-trend 1.169*** 4.220*** 4.145*** 0.015* 5.954*** 2.708*** 0.162*** −1.239** −1.248** 0.008 (SE) (0.281) (0.631) (0.690) (0.005) (0.474) (0.374) (0.017) (0.412) (0.408) (0.015) FDR-adjusted P < 0.001< 0.001< 0.0010.041< 0.001< 0.001< 0.0010.0080.0080.624 AIC 9.28812.91110.4547.3909.5788.8443.08010.29010.3155.741Cumby–Huizinga χ228.10***22.42***31.26***23.12***39.83***48.27***38.91***15.79***17.31***28.62***Note. Standard errors in italics in parentheses below the corresponding coefficient. Significance levels based on raw P values. * P < 0.05, ** P < 0.01, *** P < 0.001. The FDR-adjusted P-value row applies the Benjamini–Hochberg procedure at q = 0.05 to the post-reform trend (β₁ + β₃) row across the 10 indicators. AIC = Akaike Information Criterion. The Cumby–Huizinga test confirms that the Newey–West specification with one lag adequately accounts for residual autocorrelation. Newly enrolled patients, VL numerator, VL denominator, ART numerator, ART denominator, SMZ numerator, and SMZ denominator are reported as monthly counts; VL rate, ART rate, and SMZ rate are reported as percentages Table 3Interrupted time series estimates of pre- and post-consortium trends for HIV therapy monitoring and effectiveness indicators in M City, Yunnan Province, January 2019 to July 2024ParameterNewly enrolled (y1)VL num. (y2)VL denom. (y3)VL rate (y4)ART num. (y5)ART denom. (y6)ART rate (y7)SMZ num. (y8)SMZ denom. (y9)SMZ rate (y10)β0 baseline25.441***806.412***1591.265***51.155***1656.301***1988.493***83.302***59.728***67.235***89.630*** (SE) (12.070) (160.093) (16.169) (10.516) (5.037) (0.763) (0.253) (18.684) (21.606) (2.737) β1 pre-trend1.39153.337**8.140***2.912*7.826***1.393***0.332***3.4783.6350.181 (SE) (1.822) (18.512) (2.110) (1.224) (0.735) (0.126) (0.036) (2.533) (2.822) (0.231) β2 level shift38.5423.53964.235*−4.46953.884**64.783***−0.12829.28220.3756.190** (SE) (22.901) (183.058) (31.811) (12.227) (15.955) (11.256) (0.603) (31.733) (34.137) (2.287) β3 slope chg.−0.222−49.117**−4.000−2.897*−1.873*1.315***−0.170***−4.717−4.884−0.173 (SE) (1.863) (18.532) (2.208) (1.224) (0.895) (0.405) (0.040) (2.566) (2.851) (0.231) β1 + β3 post-trend 1.169*** 4.220*** 4.145*** 0.015* 5.954*** 2.708*** 0.162*** −1.239** −1.248** 0.008 (SE) (0.281) (0.631) (0.690) (0.005) (0.474) (0.374) (0.017) (0.412) (0.408) (0.015) FDR-adjusted P < 0.001< 0.001< 0.0010.041< 0.001< 0.001< 0.0010.0080.0080.624 AIC 9.28812.91110.4547.3909.5788.8443.08010.29010.3155.741Cumby–Huizinga χ228.10***22.42***31.26***23.12***39.83***48.27***38.91***15.79***17.31***28.62***Note. Standard errors in italics in parentheses below the corresponding coefficient. Significance levels based on raw P values. * P < 0.05, ** P < 0.01, *** P < 0.001. The FDR-adjusted P-value row applies the Benjamini–Hochberg procedure at q = 0.05 to the post-reform trend (β₁ + β₃) row across the 10 indicators. AIC = Akaike Information Criterion. The Cumby–Huizinga test confirms that the Newey–West specification with one lag adequately accounts for residual autocorrelation. Newly enrolled patients, VL numerator, VL denominator, ART numerator, ART denominator, SMZ numerator, and SMZ denominator are reported as monthly counts; VL rate, ART rate, and SMZ rate are reported as percentages Interrupted time series estimates of pre- and post-consortium trends for HIV therapy monitoring and effectiveness indicators in M City, Yunnan Province, January 2019 to July 2024 Note. Standard errors in italics in parentheses below the corresponding coefficient. Significance levels based on raw P values. * P < 0.05, ** P < 0.01, *** P < 0.001. The FDR-adjusted P-value row applies the Benjamini–Hochberg procedure at q = 0.05 to the post-reform trend (β₁ + β₃) row across the 10 indicators. AIC = Akaike Information Criterion. The Cumby–Huizinga test confirms that the Newey–West specification with one lag adequately accounts for residual autocorrelation. Newly enrolled patients, VL numerator, VL denominator, ART numerator, ART denominator, SMZ numerator, and SMZ denominator are reported as monthly counts; VL rate, ART rate, and SMZ rate are reported as percentages Discussion In this single-site retrospective analysis, the implementation of a regional healthcare consortium in May 2020 was associated with substantial gains across the HIV care continuum in M City. The number of newly enrolled ART patients showed a sustained post-reform increase of 1.169 per month (P < 0.001 after FDR adjustment), the overall ART coverage rate gained 0.162% points per month (P < 0.001), and the SMZ usage rate showed an immediate level increase of 6.190% points (P < 0.01). The overall VL test completion rate also improved, although the slope gain was small in monthly terms. Because this is a single-site study without a contemporaneous control, these findings are most appropriately interpreted as evidence of the consortium’s contribution to observed improvements rather than as sole attribution. Optimising HIV care through regional medical collaboration in underserved areas Regional medical collaboration has emerged as a practical strategy for improving HIV/AIDS care among high-risk populations in mountainous and peripheral regions. After the consortium was established in M City, the volume of patients undergoing VL testing increased substantially, although the overall completion rate gained more modestly because the eligible denominator also expanded as the registered patient population grew. Reliable VL monitoring depends in part on platform sensitivity to circulating HIV-1 subtypes [25], and sustained scale-up of VL testing has been associated with earlier identification of treatment failure and reduced selection of resistance mutations [26]. This pattern is consistent with the trajectory observed at the national level, where the test-and-immediately-treat policy and the gradual transition from CDC-led to hospital-based service models have driven similar shifts in service indicators [3]. It is useful to compare the M City experience with two often-cited international examples. Between July 2015 and January 2017, a continuous quality improvement intervention in seven primary care clinics in rural South Africa expanded antenatal HIV viral load monitoring and repeat testing, with measurable gains in timely diagnosis and infant outcomes [27]. The South African setting differs from M City in important ways. It operates within a single national health service across relatively homogeneous primary care clinics, whereas M City links a tertiary hospital, ten township health centres, and 103 village health rooms across mountainous terrain. The U.S. experience offers a different contrast. Nance and colleagues reported a rise in annual viral suppression from 32% in 1997 to 86% in 2015 across a multi-site cohort [28], although population-level suppression rates have remained more variable than the cohort figures suggest [29], and progress toward the national Ending the HIV Epidemic targets has been uneven across jurisdictions [30]. The U.S. care continuum operates on an insurance-based system in which patients self-navigate to providers, whereas the Yunnan model is distinctly place-based and government-coordinated, with the village doctor functioning as the first point of contact and the consortium acting as the coordinating layer. Several Yunnan-specific features shape the relevance of the M City findings. Cross-border movement with Myanmar contributes to imported drug use and shifts in transmission networks, particularly along the western border. Language and literacy barriers among the Jingpo, Dai, Lahu, and Wa communities reduce the effectiveness of standard health communication materials, which informed the consortium’s investment in village-level outreach. The mountainous geography itself imposes long travel distances to the city centre, so anchoring monitoring in the village and township tier shortens the testing-to-result loop and reduces patient drop-off between testing and treatment. Service delivery dynamics during the 2020–2022 pandemic period in China also showed how fragile the HIV care continuum can be in the absence of integrated structures [24], which strengthens the case for consortium-style coordination in border counties. Within the consortium, the regional CDC and community health programmes integrated patient information into a centralised system, enabling adaptive adjustments to the regional HIV strategy. Village health offices and township health centres captured potentially infected populations in remote highland villages, referred them through clearly defined channels for VL testing, and onward to Q Hospital for confirmatory care [31]. Advancements in antiretroviral therapy Antiretroviral therapy has evolved rapidly over the past decade, with broader access, better tolerability, and longer dosing intervals reshaping the standard of care [32]. Rapid and effective ART plays a pivotal role in achieving sustained viral suppression [31]. After the consortium was established, the ART numerator increased by approximately 53.9 patients in the immediate period, and the overall ART coverage rate continued to climb across the post-intervention window. Centralising patient management at Q Hospital, in alignment with the 2024 Chinese HIV diagnosis and treatment guidelines, allowed individualised regimen adjustments based on disease progression. Long-acting cabotegravir and rilpivirine intramuscular injections have emerged as transformative therapies that reduce administration frequency from daily doses to once monthly or bimonthly [33], and although these agents are not yet routine in M City, the integrated patient roster places the local programme in a position to adopt them as availability permits. For patients who experience periodic VL elevations, structured re-engagement and adherence support achieve viral resuppression in the majority of cases without regimen change [34], and the consortium’s referral pathway facilitates this kind of timely follow-up. Antiretroviral drugs are increasingly used preventatively among high-risk, HIV-negative individuals, with promising outcomes [35]. Health education programmes targeting substance use stigma have improved adherence and uptake of preventive measures, which is particularly relevant for parts of the M City catchment with a history of injecting drug use. In addition to suppressive ART, prophylaxis against opportunistic infections remains a cornerstone of comprehensive HIV care, particularly given the persistent global burden of advanced HIV-associated conditions such as cryptococcal meningitis [36]. The consortium’s promotion of SMZ uptake operationalises this prophylactic principle for the M City catchment, with the SMZ usage rate reaching 98.9% in the post-reform period. Public health implications Three concrete implications follow from this analysis. First, the consortium model offers a workable template for HIV care integration in resource-constrained border counties of southwestern China and potentially other Greater Mekong subregion settings with comparable health-system structure, including parts of northern Myanmar, northern Laos, and northern Vietnam. Translating the model elsewhere would require attention to the specific epidemic profile, the existing referral network, and the readiness of village-level providers. Second, anchoring viral load monitoring in the village and township tier shortens the testing-to-result loop in mountainous geography and supports progress toward the second and third UNAIDS 95-95-95 targets. The increase in the VL testing numerator observed in M City reflects this practical benefit, as does the parallel growth in the ART coverage rate. Cost remains a critical determinant of last-mile VL scale-up [37], and the consortium structure permits resource sharing across the network in ways that a stand-alone tertiary referral system cannot. Third, the policy combination of a shared electronic record, a single designated treatment centre, and clearly incentivised referral pathways may be considered by health administrators in other regions designing similar consortia. The integration of the District Hospital of Traditional Chinese Medicine and the Maternal and Child Health Centre in the M City model also illustrates how broader services such as prevention of mother-to-child transmission can be embedded into a single coordinated structure rather than running in parallel. The integration of Ukrainian HIV-positive refugees into the Czech health system [4, 38] illustrates the broader applicability of integrated approaches in different humanitarian and health-system contexts, even though the underlying drivers differ from those in the M City catchment. Sensitivity analyses Three sensitivity analyses examined the influence of macro-level confounders on the principal findings. Excluding the strict national lockdown months (February to April 2020) and the post-restriction infection wave months (December 2022 to February 2023) preserved the direction, magnitude, and significance of all primary results. Adding a binary COVID-19 disruption indicator into the segmented regression model produced negligible change in the post-reform slope estimates for the main indicators. Adding a step term marking the May 2022 implementation of the updated national HIV diagnosis and treatment guideline likewise left the consortium effect intact. Detailed results are presented in Supplementary Table S2. These sensitivity analyses do not eliminate the possibility of unmeasured confounding, but they suggest that neither the pandemic disruption nor the 2022 guideline update on its own can account for the observed pattern. The robustness of the post-reform slope to these adjustments is consistent with the interpretation that the consortium contributed materially to the observed improvements. Limitations and future directions Several limitations warrant attention. First, this is a single-site study without a contemporaneous control, which limits the strength of the causal inference that can be drawn. The findings should be read as evidence of contribution rather than sole attribution. Future work should adopt a controlled ITS design using neighbouring prefectures within Yunnan that did not adopt the consortium model during the same period as comparison sites, or a synthetic control approach drawing on aggregate provincial HIV surveillance data. Second, although we focused on indicators central to the WHO Consolidated Guidelines and the UNAIDS 95-95-95 framework, broader outcome measures such as cohort retention, mortality, and viral suppression rates were not analysed. As the regional information system matures, these endpoints should be incorporated. Third, the analysis was based on monthly aggregate data, which does not fully capture seasonal variation. Future studies could use weekly resolution data to characterise short-term dynamics, particularly around the lunar new year migration window when cross-border movement intensifies. Fourth, despite the sensitivity analyses described in Sect. 4.4, residual confounding from the COVID-19 pandemic and from continuing changes in HIV treatment technology cannot be ruled out. Continued monitoring of the consortium’s performance through the next five years would help separate sustained programme effects from transient post-implementation gains. Finally, the study examined a single border city, and direct extrapolation to non-border or non-mountainous prefectures should be approached with caution. External validation through multi-site or controlled designs is required before the model can be recommended for broader scale-up. Looking forward, several technological developments may further strengthen the consortium’s capacity. Point-of-care HIV viral load assays, including whole-blood platforms designed for low-resource settings [39] and compact self-administered devices [40], could shorten the testing-to-result loop further at the village tier and reduce reliance on centralised laboratory turnaround. Future technologies for monitoring HIV drug resistance and assessing potential cure indicators [41] may eventually allow more granular individual-level treatment decisions, although they are not yet routine in district-level practice in M City. Conclusion The findings from M City are consistent with a meaningful contribution of the regional healthcare consortium to improvements in HIV diagnosis, treatment, and prophylaxis indicators across the period from January 2019 to July 2024. Newly enrolled patients, VL test completion rate, ART coverage rate, and SMZ usage rate all showed sustained or stepwise gains that survived adjustment for COVID-19 disruption and for the 2022 national HIV treatment guideline update. The model offers a candidate framework for HIV care integration in resource-constrained border counties of southwestern China and potentially other Greater Mekong subregion settings with comparable health-system structure. Anchoring monitoring in the village and township tier, designating a single treatment centre, and tying providers together through a shared electronic record and clearly incentivised referral pathways all appear to contribute to the observed gains. Given that the analysis covers a single city, however, multi-site or controlled designs are needed before broader generalisation. Continued monitoring will help establish whether the gains observed during the first four years of the consortium can be sustained and translated into measurable improvements in cohort retention, viral suppression, and survival. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1