Dr. Boyle, we thank you for taking the time to comment on our article, and we are very pleased to see the journal’s comments space used to foster discussion. You raise some interesting concerns about the simplified algorithm we propose for determining whether a patient can start ART immediately or whether treatment should be delayed, even briefly, pending additional care or services.

Let us start by agreeing with you that the timing of dispensing drugs should certainly be flexible. Under the algorithm, dispensing is intended to be based on the patient’s responses to the readiness assessment. Any patient who indicates hesitation about immediate initiation should be “delayed,” which in this case might be a referral for additional counseling, help with disclosure, or just time to think about it, as you suggest.

Your other concern about the algorithm requiring too many procedures and offering too many reasons for delay is the main reason that we recommend evaluating it in real-world studies. Our intent with the algorithm is to “screen in” (identify) the subset of patients who have absolutely no indications that they cannot walk out with their initial supply of tablets on the same day. For these patients, the algorithm should be easy to apply. At the same time, the algorithm is intended to “screen out” those in whom therapy should probably be delayed. The “delay” called for may be for several days or weeks to treat a co-infection, or for no more than the few minutes for a additional attention from a clinician or counselor. For example, a patient taking concomitant medications may become confused as to whether ARVs should be added to her or his daily schedule, or should replace other medications. An additional brief interaction with a clinic counselor may be enough to allay concern such that the patient can still take home a supply of ARVs that day. Similarly, it will be up to the clinic to decide how soon to start ART in a patient with TB. National guidelines in many countries currently suggest 14 days, which is the reason for our criterion.

The additional screening tests you suggest may be feasible in some settings, and we imagine that the algorithm proposed here will be modified by others to take advantage of the resources available. For now, our experience indicates that these tests are not widely available at point of care.

If successful, the algorithm will screen out anyone who really should not start immediately, while screening in patients who should, and thus minimizing pre-initiation loss to follow up and treatment costs for both clinics and patients. We hypothesize that patients who can start immediately will comprise a large majority of HIV-positive patients who walk in the door, but that is a question for further research. We are looking forward to evaluating the algorithm’s effectiveness and costs, and we hope that others will find opportunities to apply and evaluate it as well.

In the meantime, we thank you again for your detailed input on the proposed algorithm and look forward to additional comments from readers.