Patient knows best: blinded assessment of nonadherence with antituberculous therapy by physicians, nurses, and patients compared with urine drug levels

Background. Adherence with therapy is a wide spectrum of behavior rather than a categorical state. While extreme nonadherence is readily apparent, it is rare compared to lesser degrees of nonadherence, which are difficult to predict. Aims. To compare the accuracy of doctor, nurse, and patient prediction of adherence with antituberculous therapy with urine isoniazid levels. Methods. A prospective, blinded clinical study was conducted, comparing adherence to antituberculous therapy as reported by patients, doctors, and nurses with urine isoniazid levels. We studied 173 patients with active tuberculosis (TB) recruited over 3 years in two TB clinics in Victoria, Australia. Adherence was defined as six random urine isoniazid (INH) levels being >0. Blinded assessment of adherence was completed by doctors, nurses, and patients. Lid opening and closing of computerized pill bottles were measured in a random subsample. Results. Of 173 patients, the rate of nonadherence was 24% (41/173) by urine INH, 54% (93/173) by patient self-report, 11% (19/173) by doctor assessment, and 7% (12/173) by nurse assessment. The sensitivity of prediction of nonadherence was 76% for patient self-report, 24% for doctor assessment, and 19% for nurse assessment. The 10 patients who used computerized pill bottles were all (100%) noncompliant at some stage. Conclusion. Nonadherence is common and poorly predicted by doctors and nurses, even those with extensive experience in treating TB. Contrary to popular belief, patient self-report is more reliable than doctor or nurse assessment of nonadherence. As clinicians, asking patients about adherence may be more valuable than attempting to judge for ourselves.