test_correction.py

joined_df = events.join(indx_date_df.set_index('patient_id'), on='patient_id')
joined_df.indx_date = pd.to_datetime(joined_df.indx_date, format='%Y-%m-%d')

violating_joined_df = joined_df[
    (joined_df.indx_date < joined_df.timestamp) | 
    (joined_df.indx_date - joined_df.timestamp > timedelta(days=2000))
]
joined_df = joined_df[
    (joined_df.indx_date >= joined_df.timestamp) & 
    (joined_df.indx_date - joined_df.timestamp <= timedelta(days=2000))
]
joined_df = joined_df.drop(['event_description', 'timestamp'], axis=1)
joined_df


violators = violating_joined_df.groupby('patient_id')['event_id'].count().reset_index(
    name='event_count'
).sort_values(by='event_count')
violators[~violators.patient_id.isin(joined_df.patient_id)]