The Ohio Traffic Crash Report, form OH-1, is the same document no matter which agency files it. If you work with crash data — or you are just trying to read the one from your own accident — knowing how the form is organized answers most questions before you have to ask them.

The Shape of the Form

An OH-1 has three layers, and almost every confusion about crash data comes from mixing them up:

  • The crash. One per report.
  • The units. Usually vehicles. Several per crash.
  • The people. Several per unit.

One crash produces several units, and each unit produces one or more people. Flatten that carelessly and you will count a single two-car crash with four occupants as four crashes. This is the most common mistake in homemade crash datasets.

Layer One: The Crash

The crash-level fields identify and describe the event itself:

  • Report number and investigating agency
  • Crash date and time
  • County, and city or township
  • Location — route, milepost, nearest intersection
  • Road conditions, weather, and lighting
  • Number of units involved
  • Whether it was a hit-skip

The report number combined with the agency is the real primary key. The number alone is not unique — two departments can issue similar-looking numbers in the same year. Any database built on the number alone will collide eventually.

Layer Two: The Units

A unit is usually a vehicle, but it can also be a pedestrian or a cyclist, which catches people out. A pedestrian struck in a crosswalk is a unit on the report.

Per unit you get year, make and model, direction of travel, damage description, whether it was towed, insurance information, and the unit's role in the sequence of events.

Towed-away status is a useful severity signal. Officers are sometimes conservative with injury codes at the scene, but whether a car left on a truck is not a judgment call.

Layer Three: The People

This is the layer that carries the value. For each person involved:

  • Name — last, first, middle
  • Date of birth and age
  • Gender
  • Street, city, state, and ZIP as given at the scene
  • Their unit, and their role in it — driver, passenger, pedestrian
  • Injury severity

The address is the one people most often assume is unreliable. It is as reliable as whatever the person handed the officer, which is usually a driver's license, which is usually current. Not perfect. Better than most sources.

The Injury Severity Scale

If you only pay attention to one field, this is it. Ohio uses the standard five-level scale:

  • Fatal
  • Suspected Serious Injury
  • Suspected Minor Injury
  • Injury Possible
  • No Apparent Injury

The word suspected is doing real work in those labels. The officer is not a physician and is making a field assessment minutes after a crash, sometimes in the dark, on the side of a road. It is an observation, not a diagnosis.

The practical consequence: people coded "Injury Possible" turn out to have serious soft-tissue injuries that presented the next morning, and people coded "Suspected Minor" turn out to be fine. Filtering only on Fatal and Suspected Serious gives you a small, clean list that misses most of the real injuries in the data. That can be the right call — just make it knowingly rather than by accident.

Delayed-onset injuries are the whole reason the scale reads the way it does. Concussion, whiplash and internal injury frequently do not present at the scene. The severity code records what the officer could see, not what a doctor found three days later.

The Narrative and the Diagram

Every report carries a written narrative and a diagram of the crash. This is the part that resists automation, because it is free text and a picture rather than fields.

It is also where fault reasoning lives. There is no structured field that says who caused the crash. The contributing-circumstances codes and the narrative together imply it, and insurance adjusters read both.

What Is Not on It

Just as important as what the form contains:

  • No reliable phone numbers or email addresses. If you have seen crash data with these, they were appended from a separate consumer database, not read off the report.
  • No medical records beyond the officer's severity code.
  • No claim outcome — the report is filed long before any claim resolves.
  • No attorney representation status.

Why Extracting It Is Harder Than It Looks

Every field above is real and consistently placed, which makes the form sound easy to parse. It is not, for three reasons.

It arrives as a PDF rather than as data. It runs to multiple pages, with people continued onto supplemental sheets. And the exact layout shifts between agencies and between versions of the reporting software, so any parser keyed to fixed positions breaks quietly the first time a form version changes.

That last failure mode is the dangerous one, because it does not throw an error. It just starts returning slightly wrong rows. Reading the document rather than parsing its coordinates is the only approach that survives a form revision.