Background Hypertension is the leading risk factor for death globally. Undiagnosed hypertension is common, but the incidence in hospitalised patients is unclear. There are calls for universal facility-based screening for hypertension among all attending patients. The hospital inpatient setting, where blood pressure (BP) is measured routinely and repeatedly, presents an ideal opportunity. However, international hypertension guidelines do not include inpatient BP thresholds for diagnostic or treatment purposes. We investigated the performance of current UK community BP thresholds for diagnosing hypertension in the hospital setting. Objectives Investigate the diagnostic performance of the current UK ambulatory BP diagnostic thresholds for systolic and diastolic hypertension in the hospital setting against the reference test of community-based ambulatory BP monitoring (ABPM). Design A prospective diagnostic accuracy study. Setting Hospital inpatients admitted to three UK centres were approached. Follow-up ABPM was delivered in the community. Participants Eligible patients were aged between 18 and 80 years, with no prior diagnosis of, or prescription for hypertension, and whose mean cumulative daytime BP was 120 mm Hg to 179mm Hg systolic and ≤109mm Hg diastolic from the 24th hour of their hospital admission. Interventions Participants received 24-hour ABPM 4–26 weeks post-discharge, as the reference test for hypertension, with UK diagnostic thresholds of an average daytime BP of ≥135mm Hg systolic and ≥85mm Hg diastolic applied. Participants found to be severely hypertensive at the ABPM fitting appointment were also considered reference-test positive but did not proceed with ABPM. Primary and secondary outcome measures The diagnostic performance of a mean daytime in-hospital BP of ≥135mm Hg systolic or ≥85mm Hg diastolic (index test) for the prediction of hypertension diagnosed on ABPM (reference test) was assessed using sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) as primary outcome measures. Additionally, we explored the accuracy of a range of alternative in-hospital systolic and diastolic BP thresholds against the same reference test. Results 351 participants were enrolled and 206 completed the study protocol. The average age of the 206 participants was 53 years, 55% were male, and 91 (44%) had daytime community hypertension on ABPM reference testing. Of 107 participants with raised in-hospital daytime BP, 59 (55%) had daytime community hypertension. When assessing the performance of the index test for detecting daytime community hypertension, sensitivity was 65% (59/91, 54% to 75%) and specificity was 58% (67/115, 49% to 67%). The PPV was 55% (59/107, 45% to 65%) and NPV was 68% (67/99, 58% to 77%), respectively. A further 45/206 participants (23%) had night-time community hypertension when assessed using European diagnostic thresholds for nocturnal hypertension (120mm Hg systolic or 70mm Hg diastolic), while 25/107 of those with raised in-hospital daytime BP (23%) had night-time community hypertension. When assessing the performance of the index test for detecting either day or night-time community hypertension, sensitivity was 62% (84/135, 53% to 70%) and specificity was 68% (48/71, 55% to 78%). The PPV was 79% (84/107, 70% to 86%) and NPV was 48% (48/99, 38% to 59%). Conclusions Undiagnosed hypertension is common in hospitalised patients, particularly those with raised in-hospital BP. While in-hospital BP alone is an imperfect predictor and should not be used as a stand-alone diagnostic test, this could serve as a trigger for further assessment of BP in the community after discharge.