A recent literature contends that the returns to research in medicine are declining. This conclusion is premised on two empirical findings: the quantity of medical research is rising, while improvements in health outcomes are stagnant. We argue that the first finding is an artifact of conceptual and empirical problems in the measurement of research. We propose a method for annotating unstructured text based on fine-tuning a large language model. We apply this method to construct a census of published clinical trials. We argue that clinical trials function as a natural proxy for productivity-relevant research in medicine. In our data, the quantity, quality, and composition of clinical trials have been stable since 2010. We find that measured increases in the total quantity of medical research are driven by shifts in the composition of scientific publications that do not correspond to real increases in the quantity of productivity-relevant research. We document large increases in the growth of non-trial publications, driven by publications from China and publications that review existing literature. This increase in quantity is coincident with a decline in various measures of average publication quality. Taken together, we find no support for declining medical research productivity over this period.
