How has the Covid-19 pandemic affected wheelchair users? Time-series analysis of the number of railway passengers in Tokyo

The COVID-19 pandemic led to rapid global spread and a WHO pandemic declaration on 11 March 2020. Measures such as social distancing, mobility restrictions, telework, and e-learning drove large declines in public transport ridership and modal shifts away from transit. Concerns arose that reduced public transport usage would exacerbate inequities for vulnerable populations, including people with disabilities (PwDs), who also faced new sanitation and infection-related barriers. Prior studies suggested significant impacts on PwDs’ access to services and social participation, but quantitative evidence is limited due to heterogeneous disability characteristics and small sample sizes. This study focuses on wheelchair users, who face pronounced physical barriers in transit, and asks: How has the COVID-19 pandemic affected the number of wheelchair users accessing railway services in Tokyo? Using rail passenger time series from April 2012 to December 2021 at nine stations, the study removes seasonal components to isolate pandemic effects and compare impacts on wheelchair users versus all passengers.

The authors set four criteria for relevant prior work: (a) research on PwDs during COVID-19, (b) comparative research including people with and without disabilities, (c) transportation-focused analyses during COVID-19, and (d) time-series analyses during COVID-19. They categorize prior studies into six groups: (1) qualitative examinations of PwDs’ experiences during COVID-19 (e.g., literature reviews, questionnaires), meeting criterion (a); (2) comparative studies of people with and without disabilities focusing on social and behavioral responses, meeting (a) and (b); (3) qualitative work on transport access challenges for PwDs and operator responses, meeting (a) and (c); (4) qualitative comparisons of travel behavior and community living for PwDs vs. general population, meeting (a), (b), and (c); (5) time-series analyses of transit ridership and mobility during COVID-19 for general populations, meeting (c) and (d) but not focused on PwDs; and (6) empirical analyses using paratransit records, informative for PwDs but limited for comparisons with non-disabled riders, meeting (a), (b), and (c). No category met all four criteria simultaneously. The present study addresses this gap by conducting a time-series analysis comparing wheelchair users and all passengers, thereby integrating (a)–(d).

Setting and data: The study analyzes nine train stations in the Tokyo area on five major lines connecting central Tokyo and suburbs. The period spans April 2012–December 2021, using monthly data expressed as daily averages. Wheelchair user counts are derived from station staff records of ramp board deployments (each boarding/alighting counted), provided by East Japan Railway Company (JR East). Limitations in this source include potential oversight or recording errors and exclusion of wheelchair users who did not request assistance; however, the authors argue such errors are limited and that unassisted wheelchair boarding is rare at the study stations due to remaining step/gap barriers. All passenger counts are estimated from smart card ticket gate exits. To align series and reduce confounding, the study excludes stations with transfers within ticket gates and those with new barrier-free upgrades since 2018 (e.g., central Tokyo stations prepared for the Tokyo Olympics/Paralympics). To protect privacy, wheelchair counts were aggregated across stations. Modeling approach: The authors build state-space time-series models decomposing the observed series y into components: long-term trend (t), seasonal (s), stationary AR component (p), and noise (w) for the pre-COVID period. Transformations to state-space form follow Kitagawa (2020), with parameters estimated via Kalman filtering and fixed-interval smoothing. Model orders (trend, seasonality, AR) are selected using AIC with period f=12 for monthly data. For the post-COVID model, the seasonal component estimated from the pre-COVID model is subtracted to obtain seasonally adjusted series, reflecting structural changes due to mobility restrictions that invalidate recomputation of seasonal patterns post-2020. Change-point detection: To determine the structural break between pre- and post-COVID regimes, the study applies a locally stationary AR model to seasonally adjusted data, evaluating candidate change points by minimizing the sum of AICs (AICdiv) over two subintervals for AR orders 2–10. Difference-in-differences (DiD): To avoid the structural change period around early 2020, the study compares two periods: 2019 (Jan–Dec) and 2021 (Jan–Dec), using log-transformed series for wheelchair users (intervention) and all passengers (control). Parallel trends are supported in logs before the pandemic. Counterfactual inference: A pre-COVID model trained on data up to Dec 2019 generates counterfactual forecasts for Jan 2020–Dec 2021 assuming no pandemic. The causal effect is the difference between observed and predicted values, summarized as percentage point deviations. Post-COVID predictive performance is evaluated by fitting models on Apr 2012–Mar 2021 and forecasting Apr–Dec 2021 with standard error bands.

• Both all-passenger and wheelchair-user series exhibit strong seasonal patterns pre-COVID; seasonal variation is larger for wheelchair users (≈19 percentage points across the year: +9.9%pt in May, −8.8%pt in January) than for all passengers (≈6.7%pt: +4.0%pt in April, −2.7%pt in February). • Change-point analysis on seasonally adjusted data indicates a structural break in March 2020 (minimum AICdiv at AR order 2), one month before Japan’s first state of emergency (April 2020). No additional change points are detected through December 2021. • Observed trends: All passengers rose from ~300,000 to ~450,000 (monthly daily average) pre-COVID, dropped to ~250,000 in May 2020, and partially recovered to ~360,000 by Dec 2021. Wheelchair users rose from ~30 to ~80, dropped to ~16 in May 2020, and recovered to ~40 by Dec 2021. • Difference-in-differences (2019 vs 2021): All passengers decreased by ~20 percentage points, wheelchair users by ~46 percentage points; the differential impact is ~26 percentage points more decline for wheelchair users. • Counterfactual estimation (Jan 2020–Dec 2021): Average decline vs. predicted no-COVID scenario is ~−21 percentage points for all passengers and ~−44 percentage points for wheelchair users, a ~23 percentage point greater impact on wheelchair users. The largest monthly declines occurred in May 2020 (all: −41%pt; wheelchair: −78%pt). By March 2020, wheelchair use had already fallen ~43%pt more than overall. As of Dec 2021, all passengers remained ~−18%pt, wheelchair users ~−35%pt below counterfactual. • Post-COVID forecasting accuracy: Most predictions lie within ±1 standard deviation, with larger errors during later emergency declarations (e.g., May 2021), indicating unmodeled policy effects.

The results show an early, pronounced structural change in March 2020, preceding formal emergency measures, suggesting that risk perception and fear of infection rapidly altered travel behavior. Wheelchair users were disproportionately affected relative to all passengers, with larger seasonal swings pre-pandemic and substantially greater declines during the pandemic. The amplified impact on wheelchair users likely reflects stricter infection control and service limitations in welfare and healthcare settings, reduced access to specialized facilities and equipment, and fewer opportunities for social participation among PwDs. These findings quantitatively substantiate concerns that the pandemic disproportionately reduced mobility and societal participation for PwDs. The model’s adequate post-COVID predictive performance and the absence of later change points suggest a persistent but stabilized structural shift after the initial shock, albeit with episodic deviations tied to emergency declarations. The work underscores the need for inclusive transport and public health policies that safeguard essential mobility for PwDs during crises.

Using nine years and nine months of Tokyo rail data, this study quantified COVID-19’s impact on wheelchair users by decomposing and seasonally adjusting time series, identifying a March 2020 change point, and comparing wheelchair users with all passengers. Two principal conclusions emerge: (1) the most significant structural change in ridership occurred in March 2020, before formal government movement restrictions; and (2) wheelchair rail passengers experienced a substantially larger decline—about 20 percentage points greater on average than all passengers—indicating disproportionate reductions in opportunities for social participation among PwDs. These quantitative findings extend prior qualitative evidence and highlight the importance of integrating disability perspectives into pandemic mitigation and transit planning. Future work should continue longitudinal monitoring as conditions evolve (e.g., new variants or public health crises) and incorporate richer socioeconomic data to better understand heterogeneous impacts and tailor inclusive interventions.

• The pandemic context remained dynamic during the study period; new variants or future outbreaks could alter mobility patterns, necessitating ongoing observation. • Passenger data lacked socioeconomic attributes, limiting causal attribution of background factors and heterogeneity analyses. • Wheelchair user counts relied on station staff handwritten ramp-use logs, potentially subject to oversight or recording errors, though argued to be limited. • Wheelchair users who did not request assistance (and thus did not trigger ramp records) were not captured; while likely few at these stations due to persistent step/gap barriers, this may undercount usage.