Clinical trial patient recruitment is the work of finding patients who qualify for a clinical trial and inviting them to join. It is the slowest step in most trials. Newer software uses AI models to read hospitals' records and find these patients, and to draft the trial plan itself.
Every trial has a list of rules for who may join. The rules for who can join are called inclusion criteria, and the rules for who must be left out are called exclusion criteria. For example, a trial may need adults aged 40 to 70 with type 2 diabetes, whose HbA1c, a blood test of average sugar over three months, is between 7.5% and 10%, and who are not on insulin. Many trials fall behind because hospitals cannot find enough such patients in time.
The difficulty is that the answers are scattered. Age and blood test results sit in structured fields, which software can search easily. But "not on insulin" or "had a stroke last year" are often written only in a doctor's notes, as ordinary sentences. Reading such notes with software is called clinical NLP.
Say a hospital group in Kolkata has records for 2 lakh patients with diabetes. The trial's rules are entered into the recruitment software. A healthcare data platform supplies the group's records. Where a patient's older records are at another hospital, medical record retrieval fetches them. The AI model reads the structured data and the notes, and lists 1,140 patients who appear to qualify. For each one, it shows the sentence or test result behind its decision.
A research coordinator, an employee in the hospital's trial team, then checks each patient on the list. The model can be wrong, so a person makes every final decision. This way of working is called human in the loop. The coordinator removes 210 patients the model misjudged. The hospital's doctors contact the rest, and 310 come for screening.
Patient records are sensitive. In India, the use of a person's health data must follow the Digital Personal Data Protection Act, and the hospital's ethics committee must approve how patients are contacted.
The same models are also used before recruitment begins. They draft the protocol, the trial's plan, by comparing it with thousands of earlier trials, and they flag rules that would exclude too many patients. A specialist then reviews every draft. The number of patients enrolled at each hospital is reported to the clinical trial data platform.