Assessing the impact of data diversity on background incidence rates of adverse events of special interest: the component strategy from the COVID-19 Vaccine Monitoring project

Wednesday, August 28, 2024

2:00 PM – 2:15 PM CEST

Location: Conv Hall I C

Co-Author(s)

GR

Giuseppe Roberto, PhD, PharmD

PharmD, PhD
ARS Toscana, Florence
Firenze, Italy
DM

Davide Messina

Programmer
ARS Toscana, Florence, Italy, Italy
CD

Carlos E. Durán, MD, PhD

Assistant Professor
Department of Data Science and Biostatistics, University Medical Center Utrecht
Utrecht, Netherlands
Judit Riera-Arnau, MD

PhD candidate
Department of Data Science and Biostatistics, Julius Global Health, University Medical Center Utrecht (UMCU), Utrecht, Netherlands
Barcelona, Spain
PG

Patricia Garcia

Division of Pharmacoepidemiology and Pharmacovigilance, Spanish Agency of Medicines and Medical Devices (AEMPS), Madrid, Spain, Spain
MM

Mar Martín-Pérez, PharmD, PhD

Pharmacoepidemiologist
Spanish Agency of Medicines and Medical Devices (AEMPS)
Madrid, Spain
OP

Olga Paoletti

Researcher
ARS Toscana, Italy
Carlo Alberto Bissacco

Fundació Institut Universitari per a la recerca a l'Atenció Primària de Salut Jordi Gol i Gurina (IDIAPJGol), Spain
Felipe Villalobos, MD, PhD

Fundació Institut Universitari per a la recerca a l'Atenció Primària de Salut Jordi Gol i Gurina (IDIAPJGol)
Barcelona, Spain
JC

Juan José Carreras, PhD, PharmD

Vaccine Research Department, Foundation for the Promotion of Health and Biomedical Research in the Valencian Region (FISABIO - Public Health), Avenida Cataluña, 21, 46020 Valencia, Spain.
Valencia, Spain
AM

Ainara Mira, PhD

Senior biostatistician
Vaccine Research Department, Foundation for the Promotion of Health and Biomedical Research in the Valencian Region (FISABIO - Public Health), Avenida Cataluña, 21, 46020 Valencia, Spain., Spain
Patrick C. Souverein, PhD (he/him/his)

Assistent professor
Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht Institute for Pharmaceutical Sciences (UIPS), Utrecht University
Utrecht, Netherlands
FR

Fabio Riefolo, PhD

Medical Writer
Teamit Institute, Partnership, Barcelona Health Hub, Barcelona, Spain.
Barcelona, Spain
Olaf H. Klungel, PharmD, PhD

Head of Division
Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht Institute for Pharmaceutical Sciences (UIPS), Utrecht University
Utrecht, Netherlands
Miriam C.J.M. Sturkenboom, PhD

Professor
Department of Data Science & Biostatistics, Julius Global Health, University Medical Center Utrecht (UMCU), Utrecht, the Netherlands
Utrecht, Netherlands

Presenting Author(s)

Rosa Gini, PhD (she/her/hers)

Head of Unit at ARS Toscana, Florence, Italy
ARS Toscana, Florence, Italy
Florence, Italy

Slides

Background: Data sources used to estimate background rates of events of special interest (AESI) in the Covid-19 Vaccine Monitoring project were diverse in origin of coded diagnoses and diagnostic coding systems.

Objectives: To apply a standardised methodology to assess the impact of data diversity on estimates of rates of AESI

Methods: Five data sources were used, with different combinations of origin of recorded diagnoses: primary care (PC), in-hospital (H), emergency room (ER): BIFAP, Spain (PC + H), FISABIO (Spain, PC + H + ER), SIDIAP (Spain, PC + H + ER), ARS, Italy (H + ER), CPRD, United Kingdom (PC). Lists of diagnostic codes for 39 AESI were identified in multiple coding systems using VAC4EU methodology (ICD10-CM, ICD9-CM, SNOMED, ICPC) and made publicly available. All codes were tagged as narrow (specific) or possible by two medical doctors. Cumulative incidence of diagnoses codes detected in PC, H, and ER in 2019 per 100,000 persons were calculated separately. Incidence using composite algorithms (H+ER, H+PC) and using possible diagnosis was also calculated. The impact of each component was calculated

Results: We report three cases to illustrate the methodology. Incidence of Bell’s Palsy using single components was highest using PC data, with range from 9.2 in FISABIO to 42.5 in BIFAP, in H from 1.2 in ARS to 31.6 in SIDIAP, in ER from 8.9 to 28.2. The impact of combining multiple components on incidence was: H on top of PC increased incidence +203.3% in FISABIO, +60.6% in BIFAP, and +71.9% in SIDIAP. The impact of using ER on top of H was >999% in ARS, +126.9% in FISABIO and +27.5% in SIDIAP. The impact of using PC in addition to H+ER was +14.41% in FISABIO and +96.8% in SIDIAP.

Incidence of arterial thrombosis was highest using H, with range from 204.3 in BIFAP to 342.0/100,000 in ARS. Using PC diagnoses in addition to H had an impact of +21.3% on the incidence in BIFAP, +43.7% in FISABIO and +34.2% SIDIAP. Use of ER diagnoses had no impact on the incidence using H+PC, but in ARS, where PC was not available, it had a +24.3% impact on top of using H alone.

Acute disseminated encephalomyelitis (ADEM) was only diagnosed in H. Using possible codes thecodes only, the incidence increased from 0.1 to 14.4 in ARS, from 0.2 to 6.1 in BIFAP, from 0.5 to 3.1 in FISABIO, and from 0 to 6.1 in CPRD (SIDIAP did not extract this component).

Conclusions: Bell’s Palsy is underestimated if PC diagnoses are missing. Arterial thrombosis is identified mostly in hospital diagnoses, but either primary care or emergency room may contribute additional cases. ADEM is rare and cases may be identified using both narrow and possible codes, which requires validation.

This component strategy quantifies the impact of diversity in data sources and contributes to assessment of fitness-for-purpose of data sources for specific studies.