You’re essentially doing a feature transformation from side-specific labels (left/right) into a binary condition (injured vs uninjured).

Short approach

1. Combine left/right injury columns

2. Map any injury → Injured (1)

3. No injury → Uninjured (0)

import pandas as pd

# example
df = pd.DataFrame({
    "left_injury": [0, 1, 0, 0],
    "right_injury": [0, 0, 1, 0]
})

# create new column
df["injury_status"] = ((df["left_injury"] == 1) | (df["right_injury"] == 1)).astype(int)

If injuries are text-based

df["injury_status"] = df[["left_injury", "right_injury"]]\
    .apply(lambda x: 1 if "injured" in x.values else 0, axis=1)

You’re collapsing multiple related features into a single binary indicator, which simplifies modeling and improves interpretability.

This is common in ML preprocessing when the direction or side is less important than the condition itself.