HIV Testing at Japan’s Public Health Centers May Not Prevent Late AIDS Diagnoses, Study Finds

A nationwide analysis of HIV testing in Japan has delivered a sobering reality check for public health officials: the free, anonymous screening programs run by public health centers, long considered a cornerstone of early HIV detection, may not be the decisive factor in preventing late-stage AIDS diagnoses that many assumed them to be. The study, conducted by researchers at Teikyo University and published in BMC Public Health, examined nine years of data from all 47 Japanese prefectures and found that the apparent protective association between public-health-center screening and timely diagnosis largely evaporated once stable regional differences were taken into account.

The research team, led by Yusuke Yoshino and Yoshitaka Kimura of Teikyo University School of Medicine’s Department of Microbiology and Immunology, set out to answer a question that has lingered in Japanese HIV policy circles for years. Public health centers offer free and anonymous HIV testing, a service that distinguishes them from much of the private testing landscape, yet they represent only one component of the national testing system. Whether the volume of testing conducted at these centers actually translates into fewer people being diagnosed only after their infection has progressed to AIDS has been difficult to establish, because the same regions that invest heavily in public testing also tend to differ in countless other ways from regions that test less.

To untangle this question, the researchers assembled a prefecture-level panel dataset covering 2015 through 2023, yielding 412 prefecture-year observations. Their outcome measure was the number of AIDS-stage diagnoses among newly reported HIV and AIDS cases, drawn from national surveillance data. The primary exposure was the public-health-center HIV screening rate per 100,000 residents, lagged by one year to allow time for testing to influence detection patterns, and expressed per doubling to make the effect estimates interpretable. They fitted binomial regression models that included a calendar-year trend and an indicator for the pandemic period spanning 2020 to 2023, with standard errors clustered by prefecture to account for repeated observations within each region.

The raw numbers tell a dramatic story of pandemic disruption. In 2019, public health centers across Japan conducted 103,082 HIV screening tests. One year later, that figure had collapsed by 65.0 percent to just 36,056 tests, as COVID-19 restrictions, staffing pressures, and public avoidance of medical facilities took their toll. Over the same window, the national proportion of HIV cases diagnosed at the AIDS stage climbed from 26.9 percent to 31.5 percent, a shift that immediately raised alarms among epidemiologists who view late diagnosis as both a clinical tragedy for the individual and a public health hazard, since people unaware of their infection may transmit the virus for years.

At first glance, the pooled analysis seemed to vindicate the screening programs. Across the full study period, a doubling of the lagged public-health-center screening rate was associated with lower odds of AIDS-stage diagnosis, with an odds ratio of 0.829 and a 95 percent confidence interval of 0.773 to 0.889. In other words, prefectures where public testing was more intensive appeared to catch infections earlier, before they could progress to the severe immunological deterioration that defines AIDS. Had the analysis stopped there, the conclusion would have been straightforward: expand public testing and late diagnoses will fall.

But the researchers pushed the analysis further, and this is where the study’s most important contribution lies. When they added prefecture fixed effects to the model, a statistical technique that effectively compares each prefecture against itself over time and thereby strips out all time-invariant regional characteristics, the association weakened dramatically. The odds ratio moved to 0.948 with a confidence interval of 0.866 to 1.038, a range that includes the null value of one. The apparent protective effect, in other words, could not be distinguished from what would be expected by chance once stable differences between prefectures were accounted for. Regions with robust public testing infrastructure may also have stronger healthcare systems, higher health awareness, better-funded outreach, or demographic profiles that independently promote earlier diagnosis, and the fixed-effects model absorbs all of these confounders.

The pandemic period produced equally nuanced findings. In the fixed-effects model, the 2020 to 2023 indicator narrowly reached statistical significance, with an odds ratio of 1.210 and a confidence interval of 1.003 to 1.458, hinting that the pandemic years were associated with elevated odds of late diagnosis even after adjusting for screening rates and secular trends. Yet this estimate proved fragile. When the researchers split the pandemic period into two sub-periods, 2020 to 2021 and 2022 to 2023, the association did not hold up, suggesting that the initial signal may have reflected the acute disruption of the early pandemic rather than a sustained effect, or that it was sensitive to how the analysis window was defined.

The authors are careful about what these results do and do not demonstrate. Their conclusion is that stable regional characteristics and broader pandemic-related changes may contribute to differences in AIDS-stage diagnosis, but that the evidence does not support a causal effect of public-health-center screening or of the COVID-19 pandemic itself. This is a methodologically honest position, and it carries real implications. As an ecological panel study, the analysis operates on aggregated prefecture-level data rather than individual patient records, which means it cannot trace whether any specific person who tested at a public health center avoided a late diagnosis. Ecological designs are vulnerable to what epidemiologists call the ecological fallacy, in which group-level associations fail to hold at the individual level. The one-year lag and the fixed-effects structure mitigate some concerns, but unmeasured time-varying confounders, such as changes in pre-exposure prophylaxis uptake, shifts in testing behavior among men who have sex with men, or evolving patterns of private and self-testing, could still influence the results.

The findings arrive at a moment when Japan’s HIV landscape is changing. The country has historically lagged behind many high-income nations in HIV testing coverage and in the adoption of pre-exposure prophylaxis, and late diagnosis rates have remained persistently higher than targets set by international health bodies. The 65 percent collapse in public-health-center testing during 2020 demonstrated how fragile this testing infrastructure is when confronted with a competing health emergency, and the concurrent rise in the AIDS-stage diagnosis proportion offered a natural experiment in what happens when routine screening contracts. Yet the study suggests that simply restoring or expanding testing volumes at public health centers may not, by itself, resolve the late-diagnosis problem, because the underlying drivers of delayed detection appear to be woven into the fabric of regional healthcare systems rather than into testing throughput alone.

For policymakers, the message is twofold. First, the study underscores the value of rigorous quasi-experimental methods in public health evaluation: associations that look compelling in pooled data can dissolve under fixed-effects scrutiny, and interventions should be evaluated with designs capable of separating program effects from regional context. Second, it points toward the need for a broader strategy that goes beyond public-health-center testing, encompassing community-based testing, normalization of routine HIV screening in clinical settings, expanded access to pre-exposure prophylaxis, and targeted outreach to populations at highest risk. The researchers used only aggregated, publicly available data and received no specific funding for the work, and they note that the pandemic-period association warrants continued monitoring as post-pandemic testing patterns stabilize. Whether Japan can drive down its late-diagnosis rate will likely depend less on any single testing channel and more on how comprehensively it can rebuild and diversify its HIV detection ecosystem in the years ahead.

Subject of Research: The association between public-health-center HIV screening coverage and AIDS-stage diagnosis in Japan before and during the COVID-19 pandemic

Article Title: Public-health-center HIV screening coverage and AIDS-stage diagnosis in Japan before and during the COVID-19 pandemic: a nationwide prefecture-level panel study

Article References: Public-health-center HIV screening coverage and AIDS-stage diagnosis in Japan before and during the COVID-19 pandemic: a nationwide prefecture-level panel study. (n.d.). https://doi.org/10.1186/s12889-026-29616-y

Image Credits: AI Generated

DOI: 10.1186/s12889-026-29616-y

Keywords: HIV, AIDS, public health centers, COVID-19, Japan, HIV screening, late diagnosis, epidemiology, ecological study, public health surveillance, prefecture fixed effects, Teikyo University