Purpose: To compare two methods for identifying depression among health plan enrollees with diabetesadministrative data base codes versus natural language processing (NLP) of office note verbal narratives. Background: In research studies in health care systems, depression cases can be identified through coded data in the health plan administrative data system, including diagnosis (ICD-9) codes and pharmacy codes. However, depression is often under-coded. Clinician notes offer an alternative source for identifying depression. Natural language processing tools can be used to enhance case identification where clinician notes are available in electronic form. Methods: The sample was comprised of 2680 health plan enrollees that had received a new diabetes diagnosis within the previous two years. Using phrases that included the embedded term ‘depress’, two reviewers coded all cases with an appropriate indication of depression, excluding inappropriate uses of the term ( ‘denies depression’; ‘depressed fracture’; ‘no history of depression’). On cases where there was disagreement, the two reviewers looked at the full note to determine the appropriate decision. The depression cases identified through the administrative database were compared to those identified through NLP in terms of clinical and demographic characteristics. Results: Of the 2680 incident diabetes patients, 260 (10%) were identified through both methods, and an additional 141 were identified with depression through NLP only. There were 37 depression cases identified through administrative codes with no NLP evidence of depression. There were no significant differences between NLP versus administrative data criteria in terms of gender, age, comorbidity, or A1c of those identified with depression. However, patients identified through the administrative system had more primary care visits and had longer continuous enrollment in the health plan. Conclusions: Use of NLP increased numbers of identified depression cases by almost a third. The NLP tool provides an efficient way to enhance the identification of depression cases, both for research and population-based quality improvement efforts. As more and more medical providers establish paperless offices, NLP will be an increasingly useful and efficient tool that can be used by both clinicians and researchers for a variety of purposes. This study was funded by NIH/NIDDK: Grant #DK066050.