question about eicu in drug recommendation task

[azusakou]

Thank you so much for your work! when I use eicu data for drug recommendation, I meet an error as: Key drugs has mixed nested list levels across samples.

could you please tell me how to solve this problem?

Thanks in advance

[ycq091044]

Yes, I'd like to help. Could you share with us what functions you have called?

[azusakou]

Yes, I'd like to help. Could you share with us what functions you have called?

Thanks for your reply.

from pyhealth.datasets import eICUDataset

base_dataset = eICUDataset(
    root="./eicu-crd/2.0",
    tables=["diagnosis", "medication", "physicalExam"],
    dev=True,
    refresh_cache=False,
)
sample_dataset = base_dataset.set_task(task_fn=drug_recommendation_eicu_fn)
sample_dataset.stat()
print(sample_dataset.available_keys)

the error is from: sample_dataset = base_dataset.set_task(task_fn=drug_recommendation_eicu_fn)

[ycq091044]

Thanks for your patience. I just found the issue. We have a typo in drug_recommendation_eicu_fn in the current package version. Could you please use the following function to replace (not import from pyhealth.tasks, but directly define the function and use it):

def drug_recommendation_eicu_fn(patient):
    samples = []
    for i in range(len(patient)):
        visit = patient[i]
        conditions = visit.get_code_list(table="diagnosis")
        procedures = visit.get_code_list(table="physicalExam")
        drugs = visit.get_code_list(table="medication")
        # exclude: visits without condition, procedure, or drug code
        if len(conditions) * len(procedures) * len(drugs) == 0:
            continue
        # TODO: should also exclude visit with age < 18
        samples.append(
            {
                "visit_id": visit.visit_id,
                "patient_id": patient.patient_id,
                "conditions": conditions,
                "procedures": procedures,
                "drugs": drugs,
                "drugs_all": drugs,
            }
        )
    # exclude: patients with less than 2 visit
    if len(samples) < 2:
        return []
    # add history
    samples[0]["conditions"] = [samples[0]["conditions"]]
    samples[0]["procedures"] = [samples[0]["procedures"]]
    samples[0]["drugs_all"] = [samples[0]["drugs_all"]]

    for i in range(1, len(samples)):
        samples[i]["conditions"] = samples[i - 1]["conditions"] + [
            samples[i]["conditions"]
        ]
        samples[i]["procedures"] = samples[i - 1]["procedures"] + [
            samples[i]["procedures"]
        ]
        samples[i]["drugs_all"] = samples[i - 1]["drugs_all"] + [
            samples[i]["drugs_all"]
        ]

    return samples

So, this new function is almost the same as the one that you have imported. The only difference is that in the second to last line:

samples[i]["drugs"] = samples[i - 1]["drugs_all"] + [
            samples[i]["drugs_all"]
        ]

is changed into

samples[i]["drugs_all"] = samples[i - 1]["drugs_all"] + [
            samples[i]["drugs_all"]
        ]

[ycq091044]

BTW, the task function is really just an example of how we define healthcare tasks. You can follow the pattern and modify for your own purpose.

[azusakou]

BTW, the task function is really just an example of how we define healthcare tasks. You can follow the pattern and modify for your own purpose.

It works! Many thanks!!!