Automated Telephone Outreach With Speech Recognition to Improve Diabetes Care: A Randomized Controlled Study

Randomly allocate a total of 1200 health plan members with diabetes to automated telephone
outreach with speech recognition (ATO-SR; N = 600) or usual care (N = 600). The intervention
is a series of three calls, using automated calls, originating from the health plan, using
interactive speech recognition technology, spaced approximately 4-6 weeks apart, to encourage
participants to fulfill the recommended testing (dilated eye examinations, glycated
hemoglobin, LDL-cholesterol, microalbumin) that had not been performed received in the
preceding year.

Inclusion Criteria:

- Age 18 years or older

- Diabetes, defined as either 1) filled a prescription for insulin or an oral
hypoglycemic agent; or, 2) had two outpatient or one inpatient or two outpatients
encounter claims with an ICD9-CM or CPT code indicating diabetes.

- Gap in a key diabetes management metric as evidenced by no claim for a dilated eye
examination in the prior 15 months and no claim for one or more of the following
tests: glycated hemoglobin, LDL-cholesterol, or microalbumin. (Individuals with
evidence of having received ACE-inhibitors or angiotensin receptor blockers were
considered to have had a microalbumin test.)

Exclusion Criteria:

- No primary care clinician in the data base

- Those who had previously asked the health plan to exclude them from research or
quality improvement

- Women whose claim records contained diagnoses suggesting gestational diabetes