PURPOSE: It is unknown whether data from observational studies can assist guideline developers in areas where scientific evidence is lacking. The purpose of this study is to develop a local clinical guideline for prophylaxis against venous thromboembolism based on observational study results. DESIGN/METHODOLOGY/APPROACH: The authors performed a classification tree analysis on original data from a cross-sectional study of risk factors, prophylaxis, and prevalence of deep vein thrombosis in 818 medical in-patients hospitalised in French teaching and community hospitals. The dependent variable was the use of pharmacological prophylaxis. The rate of any deep vein thrombosis detected by routine compression ultrasonography was used as an illustrative variable. Using group consensus of expert panelists, a practice guideline was developed based on the results of this study and a systematic review of the literature. FINDINGS: The resulting classification tree involved eight terminal subsets. The mean misclassification rate was 31 per cent in tenfold cross-validation and the area under the ROC curve was 0.75. The classification tree was converted into graded recommendations. According to these recommendations, 455 patients (56 per cent) required prophylaxis. The corresponding rate of deep vein thrombosis was 11 per cent (28/245, [CI = 8-16]) in 245 prophylaxis users and 11 per cent (23/210, [7-16]) in 210 prophylaxis non-users. Conversely, 363 patients (44 per cent) did not require prophylaxis: 63 actually received prophylaxis and six had deep vein thrombosis (2 per cent, [0.6-4]). RESEARCH LIMITATIONS/IMPLICATIONS: Further prospective validation of the guideline is important prior to its implementation. ORIGINALITY/VALUE: Integrating data from observational studies may be an effective way to develop guidelines when robust scientific evidence is lacking.