The relationship between heart rate variability (HRV) and hypertension is well established in multiple studies. However, there is a lack of investigation on the impact of demographics and other diseases related to cardiovascular health on the performance of HRV based hypertension detection models. This study aims to address these issues by determining the efficacy of such models in an unconstrained setting. 24 hours long ECG were recorded for 1377 subjects. HRV features from time, frequency and nonlinear domains were extracted from 1 minute long R-peak to R-peak intervals (RRIs). Demographic factors of age, gender and body mass index (BMI) were added one by one as additional features into logistic regression models. The performance of the models was analysed with respect to different age groups. The results show that inclusion of age into the HRV model increased its accuracy from 71.7% to 77.6%. However, the model’s predictions were mostly similar to the ones that would be obtained with an age based threshold. This is due to the natural age bias in the data which makes age a confounder for HRV based hypertension detection. This highlights the importance of naturally occurring demographics imbalance and how this must be carefully considered when developing HRV models for hypertension.