What this study lets us assert in 4 sentences
- Mortality is stable at a high level. Brazil has recorded about 1,300 cyclist deaths a year for more than a decade. Three independent methods (direct SIM, the WHO method on 2016, 2023 SIM) converge on the range of 1,250 to 1,380.
- The serious tier is surging. Hospital admissions grew 71% over the decade (9,238 in 2014 to 15,573 in 2023). The admissions-to-deaths ratio rose from 6.8 to 12.1. The severity pyramid is changing shape, and minor injuries remain uncounted nationally.
- The risk is concentrated in the urban cyclist. An estimated 7 to 17 million utility cyclists (base ~12 M) against 3.5 to 5.5 million sport cyclists. Urban-to-sport ratio: 2 to 3 to 1. Death happens on the road shared with cars, not on the trail.
- Policy should target the utility rider. 60% of deaths are vehicle-strike collisions. The direct annual cost to the public health system (SUS) for injured cyclists, about R$ 15 million, is a fraction of the true social cost. Making the urban cyclist visible, in the data and in policy, is the turning point.
- Why this study exists
- Question, method, and source hierarchy
- Mortality: stable at a high level
- Cross-validation by three methods
- Severity: the minor, serious, fatal pyramid
- Who dies and who gets hurt
- Urban and sport: how many there are
- The method trap: why subtracting isn’t enough
- The scissors: 12 admissions per death and rising
- Confidence map: what is solid and what is an estimate
- Implications for public policy and mobility
- Limitations and transparency
- Reproducibility
- References
1. Why this study exists
Cycling in Brazil kills about four people a day, and the number isn’t falling. Despite this, the country has an information problem: no official database classifies the crash by the purpose of the ride. The data does not tell us who crashed on the way to work and who crashed while training. Health systems record the mechanism of the crash, not the purpose of the person cycling.
IMPS lives with this gap in practice: it maintains a collaborative mapping of incidents precisely because the official record is incomplete. This study formalizes, with method, what can be asserted from what already exists, and makes explicit what still cannot be measured.
The D.Lab reading treats road safety as decision-making under uncertainty: a person chooses route, time, and equipment from a perception of risk that may or may not be calibrated to the real risk. Calibrating that perception, with auditable public data, is part of what makes an applied study deliver policy value.
- An auditable count of deaths and hospital admissions, with three independent methods that converge.
- The separation between urban and sport cyclists, with the uncertainty range made explicit and the correct algebra for combining sources.
- A reading of where policy effort should go, anchored in mechanism, victim profile, and geography.
2. Question, method, and source hierarchy
Central question. What can public data reliably assert about cyclist crash incidence in Brazil, its severity, and the split between urban and sport use?
Objectives
- Estimate cyclist mortality and morbidity and validate those counts through independent methods.
- Describe the severity of events across three levels: minor, serious, and fatal.
- Characterize who dies and who gets hurt (age, sex, geography, mechanism).
- Estimate the proportion between urban and sport cyclists.
- Document the confidence level of each finding and the remaining gaps.
How we classified the sources
A secondary-data study, with no primary collection and no submission to a journal, published open access. Every figure was classified by the reliability level of its source, and the conclusions rest only on the most robust sources.
| Tier | Confidence | Sources | Use |
|---|---|---|---|
| Tier 1 | High | SIM and SIH (DATASUS), PNAD 2015 (IBGE), PNS (IBGE), IPEA, WHO / PAHO, peer-reviewed articles | Anchor for all high-confidence claims. |
| Tier 2 | Medium | Abramet, Aliança Bike, Abraciclo, Perfil do Ciclista, fleet estimated via POF / IBGE | Triangulation and context. |
| Tier 3 | Discarded | Commercial surveys, online panels, market lists and rankings | Context at most. Never an anchor. |
The set of methods used
The reconciliation combined direct counting of microdata, application of WHO proportions, set algebra, and a consistency audit with seven independent checks. Each is described in the reproducibility section.
3. Mortality: stable at a high level
SIM records close to 1,300 cyclist deaths a year. The series shows a trough in 2016 (1,262) and a rebound to 1,381 in 2021. Between 2014 and 2024, there were 14,834 deaths. In historical perspective, cyclists’ share of road deaths rose from 1% (396 deaths) in 1998 to 4% (1,556) in 2008, and stabilized at the current level.
4. Cross-validation by three methods
The count does not come from a single source. Three independent methods converge on a narrow range:
- → Direct SIM: ~1,300 deaths/year, underlying-cause count.
- → WHO method: 3.4% of road deaths × 38,651 deaths in 2016 = 1,314.
- → 2023 SIM: direct count in the most recent reference year = 1,288.
5. Severity: the minor, serious, fatal pyramid
Looking only at deaths hides the size of the problem. An honest reading requires the severity pyramid:
- ● Fatal: about 1,300 a year (SIM).
- ● Serious (SUS hospital admissions): from 9,238 in 2014 to 15,573 in 2023, a 71% rise over the decade (SIH).
- ● Minor injuries and near misses: not measured nationally. The outpatient database does not allow isolating the road victim, so this tier, probably the largest, remains invisible in the official data.
Sources: Fatal, SIM. Serious, SIH (9,238 in 2014 and 15,573 in 2023). Minor, no consolidated national count.
The ratio is approximately 12 hospital admissions for every death, and beneath it lies a base of minor injuries that no one measures.
6. Who dies and who gets hurt
About 80% of fatal victims are men. Deaths are concentrated in the 50-to-59 age band, whereas the injured who receive care tend to be younger (peak between 25 and 34), which suggests distinct exposure profiles by age. Geographically, about 60% of deaths occur in the South and Southeast, with São Paulo leading in hospital admissions.
7. Urban and sport: how many there are
There is no official count by purpose, so the estimate is a triangulation of sources, presented as a range.
Sport, anchor: PNAD 2015 (IBGE), cycling ~9% of people who practice sport and physical activity. Urban, anchors: IPEA (7% use a bicycle as their main mode of transport) and prevalence of use for commuting measured in a population-based study.
The reading is directionally stable and gains support from the mechanism of the crashes. The urban utility cyclist, generally of lower income, is the majority and is the one who dies most, contrary to the imagery that associates the bicycle above all with leisure. Policy and infrastructure designed for the sport cyclist miss the public that most needs protection.
8. The method trap: why subtracting isn’t enough
Trying to obtain the urban figure by subtracting the sport figure from the total does not work, because the groups overlap. Someone who bikes to work and rides for leisure on weekends is in both. The correct approach is set algebra:
9. The scissors: 12 admissions per death and rising
Deaths stayed stable while hospital admissions surged. The admissions-to-deaths ratio rose from 6.8 in 2014 to 12.1 in 2023. The serious tier is growing faster than the fatal one, which is consistent with more people cycling, more exposure, without a proportional worsening in lethality per event.
10. Confidence map
Each finding was consolidated into a record with source and confidence level, and subjected to an automated consistency audit with seven checks, all passed. Commercial surveys and online panels were discarded as anchors.
| Finding | Value | Confidence |
|---|---|---|
| Cyclist deaths per year | ~1.300 | High |
| Mortality plateau over the decade | stable | High |
| Rise in admissions over the decade | +71% | High |
| Admissions-to-deaths ratio (2023) | ~12 : 1 | High |
| Deaths in vehicle-strike collisions | ~60% | High |
| Cyclists’ share of total road deaths | 3,4% | High |
| Rate per 100,000 inhabitants | 0,64 | High |
| Sport cyclists | 3.5 to 5.5 M | Medium-high |
| Urban : sport ratio | ~2 to 3 : 1 | Medium-high |
| Urban cyclists | 7 to 17 M | Medium |
| Rate per 100,000 cyclists | ~6 to 11 | Medium |
11. Implications for public policy and mobility
Campaigns and infrastructure designed for the sport cyclist do not capture the public that dies most. Policy should target utility commuting on shared roads.
60% of deaths are vehicle-strike collisions with cars. The marginal safety gain from segregation on the main road is greater than that of new bike paths in parks. Prioritize bike lanes in commuting corridors, speed reduction on shared roads, and treatment of intersections with a crash history.
As long as the official database does not classify the crash by the purpose of the ride, any national policy operates on incomplete data. Adding a purpose variable to PNAD, PNS, or the VIVA form itself would be the only cheap, structural intervention that unlocks the remaining uncertainty in this study.
Media coverage that frames cycling as a sport tends to reinforce the imagery of a middle-class hobby. The typical victim is the lower-income urban worker going to or from work. Recalibrating the framing is part of the public health effort.
The direct cost to the public health system (SUS) for injured cyclists, about R$ 15 million a year, is a fraction of the true social cost, since it does not include deaths or lost productivity. The total cost of road crashes is estimated by IPEA at R$ 28 billion a year. While the official record is incomplete, initiatives such as the IMPS collaborative mapping are a legitimate complementary source.
12. Limitations and transparency
| # | Limitation | Implication |
|---|---|---|
| 1 | Underreporting | The official numbers are a floor, not a complete portrait. There is recognized underreporting of cyclist crashes. |
| 2 | Divergent definitions | Total mortality (SIM), hospital death (SIH), and vehicle-strike subsets measure different things and cannot be added without a caveat. |
| 3 | No purpose variable | The urban-versus-sport split is indirect inference. The overlap between the two groups was not measured, and the range reflects that uncertainty. |
| 4 | Sample-based sources | Surveys such as VIVA do not generalize to the entire country. |
| 5 | Secondary data | This is a reconciliation study. It does not replace dedicated primary research (a representative survey, road instrumentation, cross-referenced microdata). |
13. Reproducibility
Every claim in this study can be reconstructed from public sources. The analysis pipeline is organized into components:
| Component | What it does | Output |
|---|---|---|
| DATASUS extraction | Pipeline for SIM and SIH microdata, isolating cyclists by ICD-10. | Annual CSVs by state and age band. |
| Mortality | Time series of deaths, trough and rebound, with a validation chart. | 2010-2024 series, trend chart. |
| Population estimate | Triangulation of PNAD 2015, IPEA, prevalence of use for commuting. | Urban and sport range, urban-to-sport ratio. |
| Reliability | SIM × WHO × SIH reconciliation, cross-validation, rates per 100,000. | Reconciliation table and confidence map. |
| Method trap | Set algebra over overlapping populations. | Venn diagram and methodological note. |
| Source hierarchy | Classification of sources by confidence level and discard rule. | Source register and tier 1, 2, 3 table. |
| Audit | Seven cross-checks of internal consistency. | Findings dossier and 7/7 audit report. |
Reference datasets: serie_obitos_ciclistas_BR.csv, estimativa_urbano_vs_esportivo.csv, reconciliacao_indicadores.csv, registro_de_fontes.csv, dossie_achados.csv. For access to the scripts and CSVs, get in touch.
14. References
ABRAMET. Associação Brasileira de Medicina de Tráfego. Survey of cyclist crashes drawn from SIM and SIH. 2020.
BRASIL. Ministério da Saúde. Sistema de Informação sobre Mortalidade (SIM) e Sistema de Informações Hospitalares (SIH). DATASUS.
IBGE. Instituto Brasileiro de Geografia e Estatística. Pesquisa Nacional por Amostra de Domicílios, PNAD 2015: Práticas de Esporte e Atividade Física. Rio de Janeiro, 2017.
IBGE. Pesquisa de Orçamentos Familiares, POF 2017-2018 (basis for the bicycle-fleet estimate).
IBGE. Pesquisa Nacional de Saúde (PNS) 2019.
IPEA. Instituto de Pesquisa Econômica Aplicada. Studies on the costs and modal share of road crashes.
WORLD HEALTH ORGANIZATION. Global Status Report on Road Safety. Geneva: WHO, 2016 and subsequent editions.
SOUSA, M. H.; BAHIA, C. A.; CONSTANTINO, P. Análise dos fatores associados aos acidentes de trânsito envolvendo ciclistas atendidos nas capitais brasileiras. Ciência & Saúde Coletiva, 2016.
TRANSPORTE ATIVO; OBSERVATÓRIO DAS METRÓPOLES. Pesquisa Nacional sobre o Perfil do Ciclista Brasileiro, 4th edition. 2024.
Aliança Bike. Sector surveys on the bicycle fleet and bicycle use in Brazil.
This study was produced by D.Lab Research in partnership with the Instituto Movimento Pedal Seguro (IMPS) as part of the open-research initiative in mobility and road safety. Want to apply this level of analysis to a decision at your organization?
Talk to D.Lab →
