Thay vào đó, chất lượng nghiên cứu sẽ giúp phát triển giả thuyết rằng sau đó có thể được đánh giá bằng cách sử dụng đánh giá bán định lượng như khảo sát. Khảo sát được khó khăn để giải thích khái quát hạn chế của họ | 4. Nonclinical Components of Surgical Decision Making 37 odologies - such as surveys case vignettes and decision-analytic modeling - all of which have important methodological Clinicians often find qualitative research . focus groups and key informant interviews difficult to interpret because the question of generalizability is more problematic and because this approach does not test hypotheses. Rather qualitative research helps develop hypotheses that may then be evaluated using semiquantitative evaluations such as surveys. Surveys are difficult to interpret because of their limited generalizability to those who respond the degree to which the question being asked is understood by the respondent and in the case of physician surveys the extent of socially normative responses. Socially normative responses occur when members of a group provide acceptable answers to questions when the real answer would generate negative social judgments. These socially normative answers can also occur in the setting of anonymous surveys but are more common when the individuals are identified. In quantitative evaluations of these issues such as in a prospective cohort that includes data on beliefs and attitudes of the surgeon and patient the number of variables of interest and potential for confounding may be overwhelming. Methods less familiar to surgeons such as the factorial experimental design may partly overcome these obstacles. Factorial design allows comparisons of differential groupings of categorical variables. For example five dichotomized variables have 32 25 unique groupings that one can analyze using hierarchical logistic regression. In essence factorial design can estimate the individual and combined effects of many variables allowing some control of confounding and may facilitate studies trying to quantify the influence of clinical and nonclinical variables. The complexity of the calculations rises with the number of variables and combinations of .
