edc-lab-results¶

Simple blood result data collection format for django models

In this design

a specimen requisition for a panel is completed first (SubjectRequisition)
result is received and entered into a result form
if a result is admnormal or gradable, an ActionItem is created.

Building the Model¶

Below we create a model class with BloodResultsModelMixin. On the class we specify the lab_panel and limit the FK the requisitions of this panel using limit_choices_to.

# models.py

from edc_lab.model_mixins import CrfWithRequisitionModelMixin, requisition_fk_options
from edc_lab_panels.panels import chemistry_panel

class BloodResultsFbc(
    CrfWithRequisitionModelMixin,
    BloodResultsModelMixin,
    BaseUuidModel,
):

    lab_panel = fbc_panel

    requisition = models.ForeignKey(
        limit_choices_to={"panel__name": fbc_panel.name}, **requisition_fk_options
    )

    class Meta(CrfWithActionModelMixin.Meta, BaseUuidModel.Meta):
        verbose_name = "Blood Result: FBC"
        verbose_name_plural = "Blood Results: FBC"

The above example has no fields for results, so let’s add some model mixins, one for each result item.

# models.py

class BloodResultsFbc(
    CrfWithRequisitionModelMixin,
    HaemoglobinModelMixin,
    HctModelMixin,
    RbcModelMixin,
    WbcModelMixin,
    PlateletsModelMixin,
    MchModelMixin,
    MchcModelMixin,
    McvModelMixin,
    BloodResultsModelMixin,
    CrfStatusModelMixin,
    BaseUuidModel,
):

    lab_panel = fbc_panel

    requisition = models.ForeignKey(
        limit_choices_to={"panel__name": fbc_panel.name}, **requisition_fk_options
    )

    class Meta(CrfWithActionModelMixin.Meta, BaseUuidModel.Meta):
        verbose_name = "Blood Result: FBC"
        verbose_name_plural = "Blood Results: FBC"

If an ActionItem is to be created because of an abnormal or reportable result item, add the ActionItem.

# models.py

class BloodResultsFbc(
    CrfWithActionModelMixin,
    CrfWithRequisitionModelMixin,
    HaemoglobinModelMixin,
    HctModelMixin,
    RbcModelMixin,
    WbcModelMixin,
    PlateletsModelMixin,
    MchModelMixin,
    MchcModelMixin,
    McvModelMixin,
    BloodResultsModelMixin,
    CrfStatusModelMixin,
    BaseUuidModel,
):
    action_name = BLOOD_RESULTS_FBC_ACTION

    lab_panel = fbc_panel

    requisition = models.ForeignKey(
        limit_choices_to={"panel__name": fbc_panel.name}, **requisition_fk_options
    )

    class Meta(CrfWithActionModelMixin.Meta, BaseUuidModel.Meta):
        verbose_name = "Blood Result: FBC"
        verbose_name_plural = "Blood Results: FBC"

Building the ModeForm class¶

The ModelForm class just needs the Model class and the panel. In this case BloodResultsFbc and fbc_panel.

# forms.py

class BloodResultsFbcFormValidator(BloodResultsFormValidatorMixin, CrfFormValidator):
    panel = fbc_panel


class BloodResultsFbcForm(ActionItemCrfFormMixin, CrfModelFormMixin, forms.ModelForm):
    form_validator_cls = BloodResultsFbcFormValidator

    class Meta(ActionItemCrfFormMixin.Meta):
        model = BloodResultsFbc
        fields = "__all__"

Building the ModelAdmin class¶

The ModelAdmin class needs the Model class, ModelForm class and the panel.

# admin.py

@admin.register(BloodResultsFbc, site=intecomm_subject_admin)
class BloodResultsFbcAdmin(BloodResultsModelAdminMixin, CrfModelAdmin):
    form = BloodResultsFbcForm
    fieldsets = BloodResultFieldset(
        BloodResultsFbc.lab_panel,
        model_cls=BloodResultsFbc,
        extra_fieldsets=[(-1, action_fieldset_tuple)],
    ).fieldsets

The SubjectRequistion ModelAdmin class¶

When using autocomplete for the subject requsition FK on the result form ModelAdmin class, the subject requsition model admin class needs to filter the search results passed to the autocomplete control.

If all result models are prefixed with “bloodresult”, you can filter on the path name like this:

# admin.py

@admin.register(SubjectRequisition, site=intecomm_subject_admin)
class SubjectRequisitionAdmin(RequisitionAdminMixin, CrfModelAdmin):
    form = SubjectRequisitionForm

    # ...

    def get_search_results(self, request, queryset, search_term):
        queryset, use_distinct = super().get_search_results(request, queryset, search_term)
        path = urlsplit(request.META.get("HTTP_REFERER")).path
        query = urlsplit(request.META.get("HTTP_REFERER")).query
        if "bloodresult" in str(path):
            attrs = parse_qs(str(query))
            try:
                subject_visit = attrs.get("subject_visit")[0]
            except (TypeError, IndexError):
                pass
            else:
                queryset = queryset.filter(subject_visit=subject_visit, is_drawn=YES)
        return queryset, use_distinct

Importing External Lab Results¶

edc_lab_results provides management commands and models to import lab results from external sources (e.g. PDF reports from a hospital laboratory) into the EDC.

Models¶

Result: Stores raw parsed lab data. Each row represents one investigation result from a PDF. Fields include patient identifiers, specimen details, timestamps, result values, units, flags, and reference ranges. A utest_id field links the result to the EDC’s internal test identifier. A subject_identifier field is resolved from the name_id on the PDF via RegisteredSubject.
InvestigationMapping: Persists the mapping between a laboratory’s investigation name (as printed on the PDF) and the EDC utest_id. Scoped by laboratory so the same investigation name can map differently at different labs. An in_reportable boolean records whether the utest_id exists in edc_reportable.NormalData.

Settings¶

Two settings control the import behavior:

EDC_LAB_RESULTS_PARSERS

A dict mapping laboratory names to dotted paths of parser callables. Each parser must accept (folder, *, tz=None) and return a pandas.DataFrame.

# settings.py

EDC_LAB_RESULTS_PARSERS = {
    "MNH": "parse_trial_labs.parse_folder",
}

EDC_LAB_RESULTS_DEFAULT_MAPPINGS

A dict of dicts providing default investigation-to-utest_id mappings per laboratory. Used as best guesses during the interactive prompt when no saved mapping exists.

# settings.py

EDC_LAB_RESULTS_DEFAULT_MAPPINGS = {
    "MNH": {
        "WBC": "wbc",
        "RBC": "rbc",
        "HGB": "haemoglobin",
        "CREATININE": "creatinine",
        "CHOLESTEROL": "chol",
        # ...
    },
}

Writing a Custom Parser¶

A parser is any callable with the signature:

def parse_folder(
    folder: str | Path,
    *,
    tz: ZoneInfo | None = None,
) -> pd.DataFrame:
    ...

The returned DataFrame must include columns matching the Result model fields. At minimum: source_file, name_id, investigation, result, units, flag, reference_range_lower, reference_range_upper, and the various datetime and specimen metadata columns.

EDC_LAB_RESULTS_PARSERS = {
    "MNH": "parse_trial_labs.parsers.parse_mnh",
    "KCMC": "my_project.parsers.kcmc_parse_folder",
}

Management Commands¶

import_labs

Parses PDF files, resolves investigation mappings interactively, and saves results to the database.

manage.py import_labs /path/to/pdf_folder --laboratory "MNH"
manage.py import_labs /path/to/pdf_folder --laboratory "MNH" --dry-run
manage.py import_labs /path/to/pdf_folder --laboratory "MNH" --output results.csv

The --laboratory flag is required. It selects the parser from EDC_LAB_RESULTS_PARSERS and scopes the investigation mappings in InvestigationMapping.

On first run, the command prompts for each unknown investigation:

Unknown investigation: CHOLESTEROL
  Best guess: chol
  Enter utest_id for 'CHOLESTEROL' [chol] or 'u' for unknown:

Accepted mappings are saved to InvestigationMapping and reused on subsequent runs. The command also checks edc_reportable.NormalData and warns about mapped utest_id values that have no normal range data.

update_mapping

Updates an existing mapping and backfills the utest_id on all matching Result rows.

manage.py update_mapping --laboratory "MNH" \
    --investigation "ABS NEUTROPHIL" --utest-id "neutrophil"

Checks for conflicts (another investigation already mapped to the same utest_id) and refreshes the in_reportable flag.