EPIDEMIOLOGICAL RESEARCH IN HOSPITAL: KEY CONSIDERATIONS FOR DESIGN AND INTERPRETATION

Abstract

Most epidemiological and observational clinical studies in Vietnam are conducted on hospitalized patients, frequently lacking a control group. Despite their practicality and data accessibility, these studies are often utilized to address questions that exceed the inferential capacity of their design. In this article, I analyze the methodological limitations of hospital-based epidemiological research, particularly in the absence of a control group, within the context of medical research in Vietnam. Drawing on conceptual analysis and epidemiological principles, the article discusses common sources of bias in hospital-based studies, including selection bias, collider bias, the lack of control groups, and the regression toward the mean phenomenon. Inpatients represent a highly selected population, influenced not only by disease severity but
also by healthcare accessibility, referral pathways, and socio-economic factors. Conditioning the analysis on a hospitalized group can distort the association between exposure and disease, creating spurious correlations or masking true causal relationships. Without a control group, observed associations are easily misinterpreted as population-level risk factors. Furthermore, before-after studies in hospitals are susceptible to regression toward the mean, leading to inaccurate evaluations of intervention efficacy. Prognostic models developed from inpatient data may perform well in training environments but often fail to generalize to broader settings.
While hospital-based research remains vital for describing severe illness, inpatient prognosis, and evaluating care processes, its inferential scope is inherently limited. Clearly identifying these methodological boundaries and anchoring interpretations to sampling conditions is essential to avoid biased conclusions and to enhance the quality of medical research in Vietnam