OBJECTIVES:  Drug strength for injectable drugs is not commonly captured in Medical or Pharmacy claims database but an important factor for understanding drug compliance, persistency and length of therapy. This study is to build a model to estimate the injectable drug strength from Insurance Claims Database. METHODS:  Total ~700k Lupron scripts from 6117 Prostate Cancer patients from the Truven MarketScan Commercial Claims and Encounters (CCAE) and Medicare Claims (MDCR) from 2000 to 2016 were used in the analysis. 35,058 scripts with drug strength information are used to build the prediction model. Both Logistic Regression and General Linear Regression model were tested. Physician specialty, type of benefit plan, geographic region of employee residence and employment status and employee classification, total drug cost and treated year are evaluated in the model. Model performance is evaluated by error rate from 10-fold cross validation. The predicted Lupron strength was also compared with the simple imputation based on total drug cost in the context of compliance rate and length of therapy. RESULTS:  Among 6117 patients treated with Lupron, 22.5mg regiment is most scripted. General Linear Regression model perform better with an error rate of 0.09. By comparing the results from no imputation, simple imputation and model based imputation, the results revealed that simple imputation is over estimating in drug strength and model based imputation is close to the actual observed data. Calculated mean of compliace rate and length of therapy for non-imputation,simple imputation and model based imputation is 89% and 18.59 months, 94% and 20.17 months, 86% and 18.38 months, respectively. CONCLUSIONS: Many injectable scripts are missing drug strength information from Insurance Claims database, using model-based prediction, drug strength can be estimated more accurately using factor such as drug total cost, treated year and other associated factors.