TL;DR
This study investigated whether common acid-suppressing medications (PPIs/H2RAs) increase COVID-19 risk, while also exploring social determinants of health. The research found no significant association between chronic acid suppression and COVID-19 infection, but did identify several medical and social factors that significantly predict infection risk.
What Problem Does This Paper Address?
The COVID-19 pandemic highlighted the need to understand risk factors for infection beyond the well-established ones like age and pre-existing conditions. Acid-suppressing medications, such as proton pump inhibitors (PPIs) and H2-receptor antagonists (H2RAs), are widely used and have been previously linked to an increased risk of other respiratory infections. This raised a question about their potential role in COVID-19 susceptibility. However, prior research on this link was conflicting. Simultaneously, the pandemic underscored the profound impact of social determinants of health (SDOH) – factors like socioeconomic status, living conditions, and access to resources – on health outcomes. This paper aimed to clarify the role of chronic acid suppression in COVID-19 risk while also systematically investigating how SDOH influence infection risk, integrating these two crucial areas of inquiry.
Methods
The study employed a case-control design, a common epidemiological approach for investigating risk factors for a disease. They selected 900 patients from three University of California medical centers who had been tested for SARS-CoV-2. The cohort was balanced with 450 positive cases and 450 negative controls.
Key methodological aspects include:
- Manual Data Extraction: Medical records were meticulously reviewed by physicians to extract information on medical comorbidities and the history of chronic acid suppression use. This ensured a high level of data accuracy and standardization.
- Geomapping for SDOH: To incorporate social determinants of health, patient zip codes were used to map publicly available data. This allowed for the integration of socio-behavioral factors at a community level.
- Statistical Modeling: A multistep logistic regression model was used to identify significant predictors of COVID-19 infection. This involved initial stepwise selection across multiple imputed datasets to ensure robustness, followed by a pooled fixed-effect model.
Datasets
The study utilized de-identified electronic health records (EHR) from three University of California Health system campuses: San Francisco (UCSF), San Diego (UCSD), and Los Angeles (UCLA).
Additionally, publicly available datasets were integrated for social determinants of health, including:
- US Census Bureau data: For geographic covariates like median household income, public transportation usage for commuting, and walking as a means of transportation.
- Community Resilience Estimates (CREs): From the US Census Bureau, defining risk factors at the census tract level.
- The New York Times Mask-Wearing Survey Data: To assess county-level mask-wearing habits.
- SafeGraph Data: For daily social distancing mobility metrics.
- Google COVID-19 Community Mobility Reports: To quantify changes in time spent at various locations compared to a baseline.
Key Insights and Ideas
- Acid Suppression is Not a Major COVID-19 Risk Factor: The most striking finding is the lack of a statistically significant association between chronic use of PPIs or H2RAs and COVID-19 infection. This provides a clearer picture than previous conflicting studies.
- The Power of Integrated Data: The study successfully demonstrated the value of combining clinical data with publicly available socio-demographic and mobility data. This approach allowed for a more comprehensive understanding of infection risk.
- Latinx Ethnicity as a Significant Risk Factor: The research highlights that individuals of Latinx ethnicity are at a disproportionately higher risk of contracting SARS-CoV-2, even after accounting for medical and socioeconomic factors. This underscores the critical role of social determinants in health disparities.
- Dementia and Obesity as Risk Factors: The study identified a higher risk of infection for individuals with a BMI ≥ 30 and those with dementia, aligning with some existing literature and suggesting potential biological or behavioral pathways.
- Protective Factors Identified: Conversely, female sex, concurrent solid tumors, and a history of alcohol use disorder were associated with a lower risk of infection. The reasons for these negative associations warrant further investigation.
Key Results
- Chronic Acid Suppression: No significant association was found between chronic acid suppression (PPI or H2RA use) and COVID-19 infection. The adjusted odds ratio was 1.04 (95% CI 0.92–1.17, P = 0.515).
- Medical Conditions:
- BMI ≥ 30 was associated with an increased risk (adjusted odds ratio 1.11, 95% CI 1.03–1.19, P=0.0054).
- Dementia was associated with an increased risk (adjusted odds ratio 1.32, 95% CI 1.09–1.6, P=0.0042).
- Non-hematogenous malignancy and history of alcoholic use disorder were associated with a decreased risk (adjusted odds ratio 0.86 and 0.81, respectively).
- Sociodemographic Factors:
- Latinx ethnicity was associated with a significantly increased risk (adjusted odds ratio 1.25, 95% CI 1.16–1.35, P=4.36e-08).
- Female sex was associated with a decreased risk (adjusted odds ratio 0.89, 95% CI 0.83–0.95, P=0.000213).
- Residing in areas with higher public transportation usage (>4.4%) was associated with increased risk (adjusted odds ratio 1.1, 95% CI 1.03–1.18, P=0.0034).
- Temporal Trends: The probability of a positive SARS-CoV-2 test decreased significantly after March, likely reflecting increased testing availability.
Limitations and Future Work
The authors acknowledge several limitations:
- Selection Bias: The study cohort consisted only of patients who were tested for SARS-CoV-2. This means individuals who were not tested, or tested less frequently, are not represented, potentially introducing bias.
- Measurement Bias: There's a possibility of discordance between electronic health record documentation and actual patient medication use.
- Ecological Fallacy: Attributing community-level data (e.g., zip code data) to individual risk can be problematic.
- Cohort Unrepresentativeness: The study focused on urban, tertiary-care academic health systems, which may not fully represent the general US population.
- Pretest Probability: Testing was performed for both symptom-triggered and pre-procedural reasons, meaning the pretest probability of infection was not uniform across all participants.
- Observational Nature: The study is observational, meaning associations identified cannot definitively prove causation.
Future work could focus on prospective validation of these findings, exploring the biological mechanisms behind the identified risk and protective factors, and investigating interventions to mitigate the disproportionate risk faced by vulnerable communities.
Summary updated