Dear Editor, Factor analysis is a scientific, time-proven robust statistical technique used in medical research to assess aspects of health and sickness.1 Most identified use of factor analysis is to mitigate “curse of multi-dimensionality” from large complex datasets consisting of numerous variables.2 By observing statistical correlations among observed variables, factor analysis can be applied to summarize multiple measurable variables to build a single specific factor also called as “construct” or “domain.” UNDERSTANDING DISEASE PATHOLOGY AND UNVEILING LATENT PATTERNS OR RISK FACTORS IN CLINICAL DATA In clinical and medical research, patient history, behavior, and symptoms are cornerstone of decision-making in diagnosis and treatment.3 However, in certain conditions, patients have vague overlapping symptoms. For instance, in clinical otorhinolaryngology, discerning sinus pathology with noninvasive imaging is expensive and requires radiological expert opinion. SNOT-22 items self-reported questionnaire followed by consultation provides reliable and validated medical advice to patients about sinonasal health and quality of life post medical or surgical intervention.4 In this communication, with the help of example of sinonasal symptoms study conducted by Upreti et al. Table 1, let’s see, how factor analysis provides tools to identify sinonasal pathology and its correlation with sleep and emotions associated variables.5 In the first step, factor analysis calculates statistical correlation between variables, factor loadings, and communalities in dataset. The observation of factor loadings and eigenvalues (more than one), guides alignment of the correlated symptoms of the SNOT-22 study in four principal components/domains namely-nasal, extra-nasal, emotional, and sleep, as shown in Table 1.5 Further interpretation of results unveils hidden variables number nine and eleven-dizziness and facial pain – as symptoms of sinus involvement and construction of the fifth domain/construct – ”sinus” which contributes 4.9% of total variance to factor model (Eigenvalue more than 1.1).Table 1: Factor matrix of the sinonasal symptoms study with sinonasal outcome test-22 item based analysis 5 Application of factor analysis to datasets involves important consideration such as sample size, data cleaning, factor rotation, type of factor analysis (confirmatory or exploratory) to yield meaningful results.2 Interpretation of identified latent variables and their relationships requires specific-domain related expertise and clinical insights to arrive at meaningful conclusions. Subgroup analysis of a portion of the data and rotation of extracted factor model should be conducted further to assess validity and reliability of measurement instruments.2,4 In the same study, by examining the structure of questionnaire in the factor matrix Table 1, researcher or clinician can evaluate whether the items intended to measure the same construct are clustering together. When items/questions show good clustering, it indicates good construct validity and reliability.5 Furthermore, factor analysis helps identify redundant or ambiguous items reported in case-records that may need to be revised or removed to improve the instrument’s/questionnaire’s metric properties. SYNTHESIZING AND VALIDATING COMPOSITE MEASURES FROM INTEGRATED METRICS IN COMPLEX DATASETS Composite measurements are used in medicine and surgery to reflect statistically complicated clinical syndromes or diseases. The construction of composite measures involves several steps, beginning with factor analysis to determine the number of latent factors and their corresponding factor loadings, indicating the contribution of each observed variable. Subsequently, scores for each latent factor are calculated, through techniques such as factor score regression or weighted summation of observed variables. The resulting composite measure serves as a summary index encapsulating the multidimensional nature of the original data. The composite measures facilitate individual or group comparisons based on a single score, particularly valuable in clinical settings for diagnosis, prognosis, and treatment outcome monitoring.1,2 In quality-of-life studies, the same SNOT-22 questionnaire may measure sinus health after surgery for example, sinonasal polyp excision.4,5 In another example Tsai et al. atherosclerotic cardiovascular disease study presents clustering of systemic inflammation markers, metabolic syndrome variables, obesity parameters, and their association with sedentary behavior and physical activity.3 These research findings can be used to modulate physical inactivity, blood pressure, diabetes, and lipid profile. Overall, factor analysis plays a crucial role in various stages of medical research, from data exploration and hypothesis generation to model development and validation, ultimately contributing to a deeper understanding of diseases, better measurement of constructs, and more effective clinical interventions. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.
Modi et al. (Sat,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: