Evaluation of a Novel Optical Smartphone Blood Pressure Application

A Method Comparison Study Against Invasive Arterial Blood Pressure Monitoring in Intensive Care Unit Patients

Olivier Desebbe; Chbabou Anas; Brenton Alexander; Karim Kouz; Jean-Francois Knebel; Patrick Schoettker; Jacques Creteur; Jean-Louis Vincent; Alexandre Joosten


BMC Anesthesiol. 2022;22(259) 

In This Article


Our results indicate that there is relatively strong agreement between BP values collected using the OptiBP™ technology and those collected with an arterial catheter for patients in the ICU. When looking specifically at which comparisons met AAMI standards, MAP and DAP values were both successful while SAP did not meet criteria. Additionally, 99% of measurement comparisons were in either risk zone A (no risk) or B (low risk) for MAP and 97% were in the risk zones A or B for SAP.

There have been three previous studies investigating the OptiBP™ technology. The first study reported that the OptiBP™ measurements were accurate and precise compared to traditional double auscultatory oscillometric BP measurements in 50 patients measured in a hypertension department.[17] The mean of the differences was -1 ± 8 mmHg for SAP, -1 ± 5 mmHg for MAP and 0 ± 5 mmHg for DAP. Importantly, very strict research conditions were used with every patient sitting and placed in the same arm position. The next study was published by our group looking at emergency room patients and it demonstrated moderate accuracy between OptiBP™ and a standard upper arm cuff.[18] In a third study,[19] conducted in patients admitted to a post-anesthesia care unit after intermediate risk surgery, there was good agreement between BP values obtained using the OptiBP™ and BP values obtained using an upper arm cuff. All these studies were performed using non-invasive reference methods that can be prone to some degree of imprecision. The current study used invasive arterial BP measurements as the reference, so additional accuracy for comparison is expected. Because of the study design (multiple time points over a two-day study period) using an invasive reference method, ICU patients were considered the most appropriate patient population, although the OptiBP™ application is not ultimately intended for ICU use.

The AAMI standards for non-invasive BP measurement state that, when comparing a new BP method with a reference method, a mean of the differences of less than 5 mmHg with a SD of ± 8 mmHg is clinically acceptable.[21] It is important to note that these standards were originally designed to assess non-automated, automated, and electronic sphygmomanometers, and not the technology used by the OptiBP™ system. Regardless, our results indicate that the OptiBP™ system met these criteria for MAP and DAP, while the SD of the SAP was unacceptably high.

Moving beyond AAIM criteria and the Bland–Altman analysis, we wanted to assess the clinical relevance of the findings in order to better facilitate potential future application. We decided that an error grid analysis with various clinical risk zones to indicate the likelihood of potential harm if the OptiBP™ values were used. At least 97% of every SAP and MAP value was located in the "no risk" or "low risk" zones, while ≤ 3% of measurements lay in the moderate risk zone (which is slightly higher than in our previous study but this is understandable as the population in the present study was more severely ill).

One notable strength of our current study is that multiple measurements at multiple times points were taken over a two-day study period in a population with expected variability in their blood pressure. However, the study also has weaknesses. Firstly, the initial calibration of the application requires a BP measurement with the reference method immediately before taking the measurements to be evaluated. As a result, all immediately subsequent BP values will most likely be quite similar to the recently recorded calibration measurement and any error in the method being tested would not be apparent. Secondly, the number of patients included in the study remains relatively small (30 patients) and the application provided measurements that could be analyzed in 73% of the included patients. It means that in 8 patients (27% of the study collective), no BP values could be measured. Reasons for inadequate signals are under investigation and may be multifactorial, such as artifacts related to poor positioning of smartphone on patient's finger, inadequate signal due to bed/patient/smartphone movements, insufficient features recognition leading to signal non-processing by algorithm or morphological anomalies on finger of patient. However, this situation was anticipated given that the population tested was receiving intensive care, which likely effects peripheral vasculature. While not directly tested, judging the performance of OptiBP™ using results from ICU patients is not ideal as this population is often in poor general health, occasionally in the prone position, and potentially receiving vasoactive medications. This clinical setting is not what the algorithm was designed, trained or intended for, but our results still provide valuable information on the performance of the application because it is compared to an invasive gold standard in a population with higher BP variability than expected in the outpatient population. Thirdly, the portability (the ability of the application to be used on multiple smartphone models and operating systems) of the application needs to be addressed. Finally, among the patients included in this study, 64 had a hypertension during the study period, which means that 37% of patients did not have the condition for which this technology was developed. Moreover, as the sample size is relatively small, no subgroup analyses could be made (hypertension vs no hypertension or diabetes mellitus vs no diabetes mellitus).

In conclusion, in ICU patients, we observed a good agreement between BP values obtained using the OptiBP™ and BP values obtained using the radial artery catheter. OptiBP™ values fulfilled the exigency of the AAMI/ISO for MAP and DAP and error grid analysis revealed that the most measurement pairs for MAP (≥ 97%) were in risk zones A (no risk) and B (low risk). However, these promising results should be taken with caution as in 8 patients, no BP values could be measured and calibration was still necessary for the remaining 22. This smartphone application will therefore need some technical improvements moving forward before any clinical integration.