sequential_drug_recommendation

[bbb801]

Dear sir/Madam,

In 'Advanced Case 2: Work on customized healthcare task' , I have questions about 'sequential_drugs' and 'drugs' (see following codes). It seems 'sequential_drugs' is always empty. What ' sequential_drugs[-1] = drugs' is used for at the final row?

def sequential_drug_recommendation(patient):
samples = []

sequential_conditions = []
sequential_procedures = []
sequential_drugs = [] # not include the drugs now
for visit in patient:

    # step 1: obtain feature information
    conditions = visit.get_code_list(table="DIAGNOSES_ICD")
    procedures = visit.get_code_list(table="PROCEDURES_ICD")
    drugs = visit.get_code_list(table="PRESCRIPTIONS")

    sequential_conditions.append(conditions)
    sequential_procedures.append(drugs)
    sequential_drugs.append([])

    # step 2: exclusion criteria: visits without drug
    if len(drugs) == 0: 
        sequential_drugs[-1] = drugs
        continue

    # step 3: assemble the samples
    samples.append(
        {
            "visit_id": visit.visit_id,
            "patient_id": patient.patient_id,
            # the following keys can be the "feature_keys" or "label_key" for initializing downstream ML model
            "sequential_conditions": sequential_drugs.copy(),
            "sequential_procedures": sequential_procedures.copy(),
            "sequential_drugs": sequential_drugs.copy(),
            "label": drugs,
        }
    )
    sequential_drugs[-1] = drugs

return samples

[ycq091044]

Thanks for the question. sequential_drug_recommendation works as an example to show how to define a healthcare task in pyhealth. By definition:

Sequential drug recommendation aims at predicting the drug set of the current visit given the history of diagnosis and procedure information as well as past drug information.

We want to use the diagnosis and procedures of visit-1 up to visit-N and the drug information from visit-1 up to visit-(N-1) to predict the drug in visit-N. A sample in this task looks like this:

{
   "patient_id": xxx,
   "visit_id": xxx,
   "sequential_conditions": [diagnoses-visit-1, diagnosis-visit-2, ..., diagnosis-visit-N],
   "sequential_procedures": [procedure-visit-1, procedure-visit-2, ..., procedure-visit-N],
   "sequential_drugs": [drug-visit-1, drug-visit-2, ..., drug-visit-(N-1)],
   "label": [drug-visit-N],
}

Some explanations for the task:

1. The current drugs are what we want to predict, so they cannot be included in the feature lists. Thus, we use

sequential_conditions.append(conditions)
sequential_procedures.append(drugs)
sequential_drugs.append([])

2. If the current drugs (i.e., predicted targets are empty, we ignore this visit).

if len(drugs) == 0: 
    sequential_drugs[-1] = drugs
    continue

3. After we obtain the sample for the current visit, we will have to add the drugs back for the next visit. Thus we have the following line in the end.

sequential_drugs[-1] = drugs

4. For the question "It seems 'sequential_drugs' is always empty." Because many patients only have one visit.

[bbb801]

Thank you very much.

…

________________________________ From: Chaoqi Yang ***@***.***> Sent: Thursday, 29 December 2022 03:54 To: sunlabuiuc/PyHealth ***@***.***> Cc: Dr. LIU Jundong ***@***.***>; Author ***@***.***> Subject: [Ext] Re: [sunlabuiuc/PyHealth] sequential_drug_recommendation (Issue #55) This email originated from outside of City University of Hong Kong. Do not reply, click links, or open attachments unless you confirm the identity of the sender and know that the content is safe. Thanks for the question. sequential_drug_recommendation works as an example to show how to define a healthcare task in pyhealth. By definition: Sequential drug recommendation aims at predicting the drug set of the current visit given the history of diagnosis and procedure information as well as past drug information. We want to use the diagnosis and procedures of visit-1 up to visit-N and the drug information from visit-1 up to visit-(N-1) to predict the drug in visit-N. A sample in this task looks like this: { "patient_id": xxx, "visit_id": xxx, "sequential_conditions": [diagnoses-visit-1, diagnosis-visit-2, ..., diagnosis-visit-N], "sequential_procedures": [procedure-visit-1, procedure-visit-2, ..., procedure-visit-N], "sequential_drugs": [drug-visit-1, drug-visit-2, ..., drug-visit-(N-1)], "label": [drug-visit-N], } Some explanations for the task: 1. The current drugs are what we want to predict, so they cannot be included in the feature lists. Thus, we use sequential_conditions.append(conditions) sequential_procedures.append(drugs) sequential_drugs.append([]) 1. If the current drugs (i.e., predicted targets are empty, we ignore this visit). if len(drugs) == 0: sequential_drugs[-1] = drugs continue 1. After we obtain the sample for the current visit, we will have to add the drugs back for the next visit. Thus we have the following line in the end. sequential_drugs[-1] = drugs 1. For the question "It seems 'sequential_drugs' is always empty." Because many patients only have one visit. — Reply to this email directly, view it on GitHub<#55 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AMPCDZM4Y7HCMRSQUA5P7FDWPSLGPANCNFSM6AAAAAATLJHX4M>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[bbb801]

Thank you. I still have a question about the following codes. May I know why 'drug' is in 'sequential_procedures.append(drugs)' instead of ' procedures'?

"""
Some explanations for the task:

The current drugs are what we want to predict, so they cannot be included in the feature lists. Thus, we use
sequential_conditions.append(conditions)
sequential_procedures.append(drugs)
sequential_drugs.append([])
"""

[ycq091044]

Good findings. This is a typo, just corrected.