← MMWR reports

mm7008e3 — Anomalies

Compared to the v-safe dataset: serious events & free-text / MedDRA gaps. Demographics out of scope. Gee J et al., MMWR 2021;70(8):283–288.

DOI: https://doi.org/10.15585/mmwr.mm7008e3

First Month of COVID-19 Vaccine Safety Monitoring — United States, December 14, 2020–January 13, 2021

Central finding #

Paper (Summary): “No unusual or unexpected reporting patterns were detected.”

This first-month safety paper markets dual-system monitoring (VAERS + v-safe). In practice the serious story is almost entirely VAERS (640 serious reports, 113 deaths, 62 anaphylaxis). From v-safe, the published Results show only solicited local and systemic reactions.

Compared to the v-safe dataset for the same window, day-0–7 check-ins include roughly 18,079 people who needed medical care, 3,222 with an emergency visit, and 450 with hospitalization—none of which appear in the Results, tables, or Summary. The same cohort also contains Death-coded free text the paper never attributes to v-safe.

Cohort: Pfizer or Moderna doses 14 Dec 2020–13 Jan 2021 with ≥1 day-0–7 check-in (~1.64M registrants). Full write-up: vsafe-analysis/mmwr/mm7008e3/anomalies_report.md.

A. Hospitalizations in v-safe — under-coded and unpublished #

Paper (Methods): “Enrollees who report seeking medical care are contacted, and a VAERS report is completed if clinically indicated.”

Checking Hospitalization is a direct structured report of inpatient care. Coders should usually attach MedDRA Hospitalisation. In this cohort there are 643 such day-0–7 responses from 450 people; only 133 (~21%) carry Hospitalisation on that response, and 166 have no MedDRA codes at all.

Free text still describes admissions (stroke workups, ICU, appendectomy, allergic observation). The MMWR never publishes these hospitalization counts from v-safe, so readers see only VAERS serious-event tallies.

Table shows an illustrative sample of responses without PT Hospitalisation (510 such responses exist in total).

B. Emergency visits — large volume, thin coding #

There are 3,538 day-0–7 ER checkbox responses in this cohort; only 581 carry an emergency-care– or hospitalisation-like MedDRA term. Most ER encounters leave no recoverable Emergency care PT even when free text explains why the person went.

Why this matters: unpublished and under-coded ER visits let “no unexpected serious patterns” rest on VAERS while active surveillance recorded thousands of emergency presentations.

Sample: ER check-ins with free text and no ER-like PT (one row per registrant).

C. Serious free text coded only as Unevaluable event #

Tens of thousands of MedDRA rows in this window are soft labels such as Unevaluable event. After filtering obvious workplace “I work at the hospital” false positives, free text still includes hospital admissions, cardiac language, and emergency evaluations sitting under weak-only coding.

D. Anaphylaxis in v-safe — unused next to VAERS #

Paper (Results): “Sixty-two reports of anaphylaxis have been confirmed…”

The paper’s anaphylaxis discussion is VAERS-framed. It does not publish v-safe MedDRA counts. In this cohort’s coding window: 213 Anaphylactic reaction, 203 Anaphylaxis treatment, and 16 Anaphylactic shock (distinct registrants). Definitions differ from VAERS “confirmed,” but this is still a large active-surveillance signal left off the page.

Table shows a sample of up to 200 registrants (213 Anaphylactic reaction / shock total).

E. Deaths reported in v-safe — omitted while VAERS deaths are reviewed #

Paper (Results): “A total of 113 deaths were reported to VAERS…”

The paper never states that deaths (or death-coded free text) were reported through v-safe. Under inclusion, MedDRA through 15 Feb 2021 includes 11 Death, 4 Cardiac arrest, 2 Foetal death, plus Sudden cardiac death / Cardiac death—18 distinct registrants in that union.

MedDRA is not a death certificate; some entries are proxy reports. Even so, free text such as “Got pulmonary embolism and died,” “Died on Feb4 from complications of stroke,” and “Passed away” were coded and never appear in the MMWR’s v-safe Results.

F. Call-center follow-up #

Paper (Methods): “Enrollees who report seeking medical care are contacted, and a VAERS report is completed if clinically indicated.”

Among 450 hospitalization reporters, GDIT final outcomes are roughly Successful 185 vs Unsuccessful 260. More than half did not receive a Successful disposition. The paper does not disclose those rates (see GDIT column on the hospital grid).

Summary #

Bottom line: Compared to the v-safe dataset, this MMWR uses v-safe as a reactogenicity brochure while medically attended care, under-coded hospital/ER events, anaphylaxis codes, and death-coded free text stay offstage—and VAERS alone carries the serious-safety reassurance.