Writing letters to patients

Writing letters to patients is a key part of my clinical practice. Henry Tufo, the mastermind behind the innovative Given Comprehensive Care Clinics at the University of Vermont, made sharing medical records with patients a standard practice years ago. Activating the patient for self-management is the goal and my patients still love getting a copy of their clinical notes, including any recommendations we discussed. What happens when we provide more information to the patient?

An article in this month's American Journal of Managed Care expands our knowledge of the effect of clinical decision support when it is extended to patients as well as providers. A team from ActiveHealth Management in New York City (part of the insurance giant Aetna), evaluated a system which sends decision support messages based on lab and billing data. The alerts to the physicians are followed 10 days later by a letter in lay language to the patient. Using an observational (non-randomized) design and claims data, they evaluated how often the alerts generated the intended action (changing a medication, ordering a test, etc.).

Overall compliance with the alerts was only middling (~30%), but the rates went up when they started sending the letters to patients. The absolute gain of 3% was a statistically significant 12.5% relative gain when adjusted for age, sex, and the mix of alertable issues. In other words, they needed to mail about 33 letters to patients to generate one additional clinical action.

How could the system be improved? The article didn't say how good the data were that the alerts were based on. Billing data are notorious slow to arrive and full of errors. Basing alerts on near real-time information from laboratories and clinical records would help.

Activating the patient sooner rather than later might improve things, too. The Vermedx Diabetes Information System mails the patient letter the same day that it sends out the alert to the provider. This gives the provider only a few days of "grace" and encourages them to develop fast and effective office systems (so they get fewer alerts in the first place and deal with them before the patient calls).

Although the authors did not present any financial analysis, my back-of-the-envelope calculations are promising: If a letter costs $1.50 to generate and mail, the cost of moving one patient to a better state of care would be about $50. This is likely a very good bargain for many conditions. What's missing, though, is convincing evidence that the system really improves outcomes and saves money. For that, we will need a large scale randomized trial.