Full datasets for my PACVS v-safe analysis

And an explanation of the schema

This is the 2nd article in the PACVS series

I have created two datasets - one for Pfizer and one for Moderna - of all the results of the LLM extraction for identifying PACVS cases in v-safe based on the Neuropathy MedDRA code.

Click here to see the full Pfizer dataset

Click here to see the full Moderna dataset

I use a Pydantic class to extract the information.

I will be referring to this schema frequently in future articles, so it is a good idea to quickly skim these fields.

The following is an explanation of the schema (I used AI to generate the explanation)

---

This document explains every field in pydantic_vsafe_neuropathy.py, including required inputs, optional inputs, and fields auto-computed by the model validator.

Field-by-Field Breakdown

Top-Level Final Assessment

• is_pacvs (CitableBool, derived): Overall classification for whether this is a clear PACVS neuropathy case.
  Derived from the heuristic score (score >= 7).

Identifiers

• registrant_code (CitableStr, required): V-safe registrant identifier.

• vaccine_manufacturer (CitableStr, required): Vaccine maker (for example, Pfizer, Moderna, Janssen).

• dose_number (CitableInt, required): Dose judged most likely responsible.

• days_from_vaccine_to_onset (CitableOptional[int], optional): Days from relevant dose to first neuropathy symptom.

Prior COVID and Baseline Health

• had_prior_covid_infection (CitableBool, required): Evidence of COVID-19 infection before symptom onset.

• seemed_healthy_previously (CitableBool, required): Whether patient appeared previously healthy without major chronic illness.

• pre_existing_conditions (CitableStr, required): Conditions that might alternatively explain symptoms.

Neuropathy Pattern Features

• neuropathy_pattern (CitableOptional[NeuropathyPattern], optional): Dominant neuropathy pattern label.

• paresthesia_tingling (CitableBool, required): Numbness, tingling, pins-and-needles.

• burning_dysesthesia (CitableBool, required): Burning pain or dysesthesia.

• facial_cranial_involvement (CitableBool, required): Involvement of face/lips/head.

• patchy_non_length_dependent (CitableBool, required): Patchy, migratory, or non-length-dependent distribution.

Multi-System Cluster

• prominent_fatigue (CitableBool, required): Significant fatigue or post-exertional malaise.

• brain_fog (CitableBool, required): Memory, concentration, or cognitive fog symptoms.

• dysautonomia (CitableBool, required): Dysautonomic features (for example tachycardia, orthostatic dizziness, palpitations).

• headache (CitableBool, required): New or worsened headache burden.

• multi_system_count (CitableInt, derived): Count of the four non-neuropathy systems above that are positive (0-4).

Impact and Chronicity

• was_debilitating (CitableBool, required): Whether daily activities/work/quality of life are substantially impaired.

• appears_permanent (CitableBool, required): Whether presentation suggests long-term disability or permanence.

• follow_up_duration_months (CitableOptional[int], optional): Approximate follow-up duration in months.

Categorization

• is_vaccine_induced_neuropathy_not_pacvs (CitableBool, required): Case appears vaccine-induced neuropathy but not full PACVS.

• is_mild_resembling_pacvs (CitableBool, required): Mild PACVS-like presentation with low confidence.

Objections and Alternatives

• other_common_causes_ruled_out (CitableBool, required): Whether common alternatives (for example diabetes, B12, thyroid) are unlikely/ruled out.

• alternative_explanations_considered (CitableStr, required): Which alternatives were considered and why less likely.

• most_likely_alternative_diagnosis (CitableOptional[str], optional): Best alternative diagnosis if PACVS is not favored.

• objective_evidence_available (CitableBool, required): Whether objective testing exists.

• symptoms_too_mild_or_nonspecific (CitableBool, required): Whether symptoms are too mild/nonspecific for strong attribution.

Why Not Regular Neuropathy

• why_not_regular_neuropathy (CitableStr, required): Reasoning against ordinary/idiopathic neuropathy.

Free-Text Importance

• key_free_text_phrases (CitableStr, required): Most important quoted free-text supporting assessment.

• why_free_text_necessary (CitableStr, required): Why coded/structured fields alone are insufficient.

Compensation Recommendation

• recommend_compensation (CitableBool, derived): Recommendation to compensate if injury is confirmed.
  Derived as True only when:

  • heuristic likelihood score is at least 8,

  • was_debilitating is true, and

  • appears_permanent is true.

• compensation_reasoning (CitableStr, required): Short justification for compensation stance.

Objection Response Summary

• objection_response_summary (CitableOptional[str], conditionally derived):

  • Populated only when is_pacvs is true.

  • Left empty (None) for non-PACVS cases.

  • Contains a concise narrative addressing common objections (timing, prior infection, pre-existing conditions, objective evidence).

Overall Assessment

• overall_pacvs_likelihood (CitableInt, derived): 1-10 likelihood score (capped at 10).

• confidence (CitableStr, required): Confidence label (for example High / Moderate / Low).

• final_conclusion (CitableStr, required): One-paragraph final case assessment.

Derived Logic (Validator Summary)

The @model_validator(mode="after") computes and/or overwrites several fields:

• multi_system_count

• overall_pacvs_likelihood

• is_pacvs

• recommend_compensation

• objection_response_summary

Heuristic score components:

• +3 if onset is clearly temporal (0-42 days after dose)

• +2 for atypical pattern signals (patchy_non_length_dependent or facial_cranial_involvement)

• +2 for multisystem involvement (multi_system_count >= 2)

• +2 if common alternative causes are ruled out

PACVS threshold:

• is_pacvs = True when total score is at least 7.

Important Implementation Note

Although some fields are marked required in the schema, the validator recomputes key derived fields after model creation. In practice, caller-provided values for derived fields may be replaced by computed values.

How to Read This Schema

• Most fields use citable wrappers (CitableBool, CitableInt, CitableStr, CitableOptional[T]).

• Each citable field typically carries:

  • value: the actual value

  • citation: one or more source reference indices

  • explanation: optional rationale text

• The model forbids unknown keys (extra="forbid"), so only the fields below are allowed.

Enum: NeuropathyPattern

Possible values for neuropathy_pattern.value:

• small_fiber

• sensory_patchy

• facial_cranial

• migratory

• length_dependent

• motor_involved