OBJECTIVES: Instrumental variable approach gained popularity in outcomes research because even in the presence of unmeasured confounding, such methods may consistently estimate the average causal effects. However, for this consistent estimation to be achieved several conditions must hold. Recent outcomes research literature seems to ignore these conditions and the results may provide inconsistent estimates. In this paper, we will overview insturumental variable approach, as well as possible tests to check the pre-requisite conditions and show how weak instruments may produce inconsistents and inefficient results. METHODS: Recently proposed Swanson and Chao tests used to check for weak instruments. For the cases with single endogenous variables conditional likelihood ratio is implemented and used. For the cases with more than one endogenous regressor Anderson Rubin test is reviewed. Relevant level curve presented for two endogeneous variables. RESULTS: The Market Scan private insurance data base was used in this study which based upon analytic sample of 36,341 asthma individuals whose health care was provided under a variety of fee-for-service (FFS), fully capitated and partially capitated health plans , including exclusive provider organizations, indemnity plans. We showed how to test weak insturements under the assumption that we have single endogenous variables. We showed that one of the insturement was weaked and the results produced inconsistent and inefficient estimates of the effect of treatment. With the correct instruments, the estimates were statistically significant (p=0.001). CONCLUSION: Despite the obvious benefit of instrumental variable models, the method should not be used blindly. Several strong conditions required for these models to work, and each of them should be tested. Otherwise the results would be statatistically worse than the results that can be achieved using simple ordinary least squares.