A convenient method to verify the accuracy of oscillometric blood pressure monitors by the auscultatory method: A smartphone‐based app

Abstract

It is recommended that oscillometric devices be calibrated by auscultation when first used, but this is difficult in practice. Here, we introduce a smartphone‐based technique to verify the accuracy of blood pressure monitors (BPMs). We enrolled 99 consecutive subjects and tested 6 brands of BPMs in this study. During measurements of electronic oscillometric BPMs, Korotkoff sounds were simultaneously collected using a stethoscope head beneath a cuff connected to a smartphone, and an app named Accutension Stetho could then yield an auscultatory BP reading as a reference. Next, differences in BP between the different BPMs and Accutension Stetho were determined. The percentage of BP differences falling within 5, 10, and 15 mm Hg; the mean (MD) value; and the standard deviation (SD) of BP differences and deflation errors were analyzed among all the BPMs. We found that the percentages of SBP differences falling within 5 mm Hg of the 6 BPMs were 80%, 79%, 77%, 72%, 68%, and 63%, in turn. The deflation rates among the 6 BPMs were 2.23, 3.48, 6.10, 2.44, 3.66, and 4.85 mm Hg/beat, respectively. Deflation errors, which were defined as deflation prior to the end of the Korotkoff sounds, existed in 4 BPMs. In conclusion, Accutension Stetho could detect BP differences between oscillometric BPM readings and simultaneous auscultatory readings. Diastolic BP was overestimated when the device deflated prior to the end of the Korotkoff sounds. Using the app, it is possible to evaluate the accuracy of BPMs among the same subjects.

1. INTRODUCTION

Blood pressure measurement based on the oscillometric method is currently a popular technique. The method works as follows: Changes in the oscillation amplitude draw the oscillometric waveform envelope (OWE), which is used for the assessment of systolic (SBP), mean, and diastolic blood pressure (DBP).

Although most oscillometric devices show good agreement with reference to the auscultatory method based on Korotkoff sounds in a vast majority of cases, agreement is remarkably poor in some cases and seems to be related to the shape of the oscillometric waveform.1 The reliability of oscillometric blood pressure measurement is especially questionable in pregnant women, children, and the elderly2 and may also be questioned in some clinical situations,3 especially when BP is either very low or very high, when patients have increased arterial wall stiffness, or when there is an arrhythmia.4

We know that the systolic and diastolic pressure values provided by oscillometry are not directly determined but estimated with undisclosed proprietary algorithms. These algorithms are why blood pressures measured by different oscillometric devices sometimes are not the same. Therefore, oscillometric devices are not interchangeable, and the same device should be used for the follow‐up of a given patient.3 Some devices use a fixed characteristic ratio for the determination of SBP and DBP, which has been shown to have substantial inaccuracies.5 The values of the fixed ratios represent population averages, so the devices may be accurate in only subjects with normal BP levels.6 Others use the OWE maximum upslope and downslope (ie, the maximum values of its derivative with respect to cuff pressure) to correspond to SBP and DBP, respectively.7 Nevertheless, both the height‐ and slope‐based techniques remain prone to noise and movement artifacts. A number of alternative techniques have also been proposed.3, 8

Additionally, results may vary greatly depending on how the oscillometric signal is processed, filtered, and differentiated.3, 9

The AHA/ASH/PCNA Scientific Statement pointed out that, even when users utilize validated devices, a successful validation test does not mean that the device will provide an accurate reading for all patients.10 There might be a substantial number of individuals among whom the error is consistently >5 mm Hg with a device that achieves a passing grade.11 The ESH standard allows 35% of readings to be >5 mm Hg. This may be more likely to occur in elderly or diabetic patients.4

For this reason, it is recommended that each oscillometric monitor should be validated on each patient before the readings are accepted.10 Once a monitor has been purchased, it is recommended that the patient bring it into the office to verify both the patient's technique and the accuracy of the device. This procedure should be repeated annually.3 Therefore, periodic validation and calibration of oscillometric devices are crucial. Furthermore, there is an urgent need to develop a convenient and accurate method to verify automatic oscillometric devices.

Some experts claim that doctors who cannot calibrate sphygmomanometers should stop measuring blood pressure.12 Under some circumstances, simulators that generate various oscillometric waveforms are used for the calibration and evaluation of automatic noninvasive BP devices, and they also help us understand the limits and pitfalls of oscillometry.13, 14

To date, no attempts to verify the accuracy of oscillometric devices have been satisfactory. Our previous study showed that Accutension smartphone auscultatory blood pressure kit (model: XYZ‐110), which used the auscultatory method to measure blood pressure, fulfilled the requirements of the ANSI/AAMI/ISO 81060‐2:2013 standard. Compared to manual auscultation, the mean device‐observer difference in 255 separate BP data pairs was 2.45 ± 2.24 mm Hg for SBP and 0.69 ± 2.09 mm Hg for DBP.15, 16 Here, we test the hypothesis that a smartphone‐based technique named Accutension Stetho, which uses the same software algorithm as Accutension XYZ‐110, can confirm the accuracy of automatic oscillometric devices.

2. MATERIALS AND METHODS

2.1. Subject selection

All the subjects in this study were Chinese participants randomly recruited from the outpatients and hospitalized patients of Shanghai General Hospital. Written informed consent was obtained from all participants. A total of 99 consecutive patients attending a routine clinical visit with their cardiologist (male 53 and female 46) were eligible to participate in this study. The overall mean age of the participants was 56.5 years (age range 21‐88 years).

The inclusion criteria were as follows: (a) males or females over 18 years of age; (b) subjects who agreed to participate in the study and completed the informed consent; (c) patients who were in a stable clinical phase. Subjects who met any one of the following criteria were excluded from the study: (a) suffering from acute pain or in a clinically unstable phase; (b) upper arm missing or upper arm wounds not healed; (c) bilateral upper arterial occlusion. The dropout criteria (withdrawal from the study midway) included the following: (a) subjects wanted to withdraw; (b) researchers believed that the subject was unsuitable to continue. The study population was a consecutive series of participants defined by the selection criteria.

All of the participants had three simultaneous, valid oscillometric and Accutension Stetho measurements for SBP and DBP readings. For more details about subject selection, see Table 1. Data were collected before the test was performed. Among the subjects, 33 subjects had hypertension, 18 had diabetes mellitus, and 21 were smokers.

Table 1.

Screening, recruitment, and subject details

Screening and recruitment		Subject details
Total screened	108	Sex
Total excluded	9	Male:female	53:46
Arrhythmias	2	Age (y)
Poor‐quality sounds	4	Range (Low:High)	21:88
Cuff size unavailable	1	Mean (SD)	56.5(16.3)
Observer disagreement	1	Arm circumference (cm)
Other reasonsa	1	Range (Low:high)	22:32
Total recruited	99	Mean (SD)	27.9(4.1)

DBP, diastolic blood pressure; SBP, systolic blood pressure.

^aExplanation summary: if any two reference SBP determinations differed by more than 12 mm Hg, or any two reference DBP determinations differed by more than 8 mm Hg.

The present study was approved by the Ethics Committee of Shanghai Jiao Tong University, Shanghai, China. Written informed consent was obtained from all participants.

2.2. Study design

The left or right arm was determined randomly. The same arm simultaneous method was employed; one subject should yield three valid measurements. Using the app installed on a smartphone, the whole measurement process was videotaped. Through playing back the video, the auscultatory readings could be determined by two observers who were blinded to each other. The video could be played back repeatedly until the values were finally confirmed. If the two observers had different determinations of SBP or DBP from the same reading, a final decision was made by a third assigned observer. Oscillometric BP values could be read from the BPMs. This specific design approach was used for two reasons: first, to minimize observer bias; second, to minimize the possibility that the oscillometric readings might affect the auscultatory readings because the observers were not blinded to the readings.

2.3. Device

Accutension Stetho is an inexpensive (<$30 US) smartphone/stethoscope kit combination that estimates precise BP values by auscultation to confirm the accuracy of an automated sphygmomanometer's readings for individual patients.17 Accutension Stetho consists of software (it can be downloaded from the App Store and installed on an iPhone) designed to verify electronic blood pressure monitors (BPMs) in clinical settings and at home (Figure 1A), and a stethoscope head equipped with a mini‐microphone used to collect the Korotkoff sounds for the smartphone. The app works as follows: simply plug one end into a smartphone, apply the blood pressure cuff, and fix the stethoscope's head below the cuff and above the artery (Figure 1B). Then, start the app, hit the “calibrate” button and start the blood pressure machine to be tested, and focus the video capture window on the machine. Next, press the “record” button, keeping the whole measurement process on the video screen within the app. Once deflation stops, hit the “record” button again—then the app is ready to determine the BP values. The recording must be played back to listen to the sounds and obtain the auscultatory readings of SBP and DBP. Once this is complete, slide your finger along the progress bar to see the exact cuff pressure corresponding to the sound that can be heard. The higher number (the one that is visible when the sound first appears) is the SBP, and the lower number (the last number visible just as the sound disappears) is the DBP (Figure 2). Touch the SAVE icon once the determined values have been entered. The measurement data including the video and the Korotkoff sounds can also be sent through email for either confirmation or further analysis.

Figure 1.

(A) The main menu of the Accutension Stetho app. (B) Demonstration of how to simultaneously perform verification of an electronic blood pressure monitor with Accutension Stetho

Figure 2.

How to determine the corresponding auscultatory readings from a BPM measurement. This picture consists of three parts. Pictured above is the video capture window. In the middle is a group of blue waveforms which represent captured Korotkoff sounds. At the bottom, the data record panel is shown. The red lines in the middle mark the first and last sounds that can be confirmed by playing back the video. Here, you can see that the corresponding BP values displayed by the BPM in the video capture window are 122 (A) and 89 (B), and you can easily recognize the auscultatory SBP as 122 mm Hg and DBP as 89 mm Hg. The oscillometric SBP and DBP values (117 and 87) can be read from the automatic BPM. When both the auscultatory and automatic SBP and DBP readings are shown in the data record panel, you can see the differences between them

In this study, we initially enrolled 6 brands of oscillometric electronic BPMs. The names and models of the 6 BPMs are as follows: Beurer BM26 (Beurer GmbH, Germany), Microlife 3MU1‐4D (Microlife USA, FL), AND UA‐771 (AND Electronics Co., Ltd, Shenzhen, China), Mabis 04‐596‐008 (Briggs Healthcare Waukegan, IL), Yuwell 670D (Jiangsu Yuyue Medical Equipment & Supply Co., Ltd, Jiangsu, China), and Omron HEM‐7051(Omron Healthcare Co., Ltd, Japan). Each BPM was used with its own cuff. Before the study started, the reading agreement between a mercury sphygmomanometer and each BPM was tested by connecting them via a Y‐tube connector.

In addition to determining the BP values, the app can be used to calculate the deflation rate of the oscillometric BPMs. The deflation rate in mm Hg/beat = (SBP‐DBP)/(n‐1) (n = numbers of Korotkoff sounds) (Formula 1). The deflation rate in mm Hg/sec = the deflation rate (mm Hg/beat)/(60/HR) (HR=heart rate) (Formula 2). SBP and DBP are determined by the auscultatory method. Heart rate can be read from the electric BPMs, since most of them have this function.

2.4. BP measurement

Subject preparation: Subjects were seated comfortably with the legs uncrossed and feet flat on the floor. The back, forearm, and elbow were supported. The middle of the cuff was positioned at the level of the right atrium of the heart.

Observer preparation: Observers were trained in using the Accutension Stetho app and the electronic BPMs.

Measurement: Measurements were performed by the automatic electronic BPMs. The appropriate BP cuff size was selected according to the mid‐arm circumference of the participant. At least three qualified systolic and diastolic measurements were expected to be obtained. The determinations were repeated at 60‐second intervals until the required number of valid determinations was obtained.

All data from a subject were excluded if any two reference SBP determinations differed by more than 12 mm Hg (1.60 kPa), or if any two reference DBP determinations differed by more than 8 mm Hg (1.07 kPa).

The mid‐arm circumference was determined by making a horizontal mark at the midpoint at the posterior aspect of the arm and measuring the arm circumference.

Three observers were involved in this study. There were also 2 supervisors who finalized the results together.

2.5. Evaluation of accuracy

We know that, although various BPMs passed validation, accuracy varied among them. It is easy to understand that the smaller the differences between the observer and device measurements, the better the accuracy. However, so far, there is no universal standard to evaluate the accuracy of BPMs. Our evaluation took several factors into account. First, as the ESH standard requires, the number of SBP or DBP differences (at most 99) between the observer and device measurements falling within 5, 10, and 15 mm Hg must all be fulfilled according to the protocol. Therefore, the greater the number of BP differences falling within 5 mm Hg, the better the accuracy. Table 2 lists the percentage of BP differences falling within different mm Hg (5, 10, and 15) for all 6 BPMs, and the percentage within 5 mm Hg was highlighted as the main evaluation index. Second, since ANSI/AAMI/ISO uses the mean difference (MD) and standard deviation (SD) to evaluate the differences between observers and devices, we also used them to estimate the accuracy (Table 3). The smaller the MD and SD, the better the accuracy. Lastly, some brands of BPMs might deflate prior to the end of the Korotkoff sounds, which inevitably leads to the overestimation of DBP.

Table 2.

Percentages of SBP and DBP differences falling within 5, 10, and 15 mm Hg

	SBP			DBP
	5 mm Hg	10 mm Hg	15 mm Hg	5 mm Hg	10 mm Hg	15 mm Hg
A	80	92	97	78	90	98
B	77	93	97	77	88	98
C	72	91	96	63	74	97
D	79	90	95	65	85	95
E	63	82	93	71	92	96
F	68	83	92	68	80	94

The data show the percentages of BP differences within 5, 10, and 15 mm Hg. The percentage of 5 mm Hg in Bold values is highlighted as the main evaluation index.

Table 3.

MD and SD of BP differences between 6 BPMs and Accutension Stetho

	SBP1		DBP1		SBP2		DBP2
	MD	SD	MD	SD	MD	SD	MD	SD
A	1.01	3.13	0.43	3.74	1.39	3.16	0.04	3.75
B	3.66	4.00	−1.93	4.21	3.42	4.00	−1.69	4.31
C	−0.93	4.94	−2.64	6.80	−2.48	4.84	−1.09	6.95
D	3.26	4.23	4.75	3.80	3.54	4.25	4.47	3.78
E	4.18	5.99	−0.74	5.31	3.85	6.11	−0.41	5.32
F	5.10	6.70	1.49	6.43	4.17	6.60	0.57	6.50

The data show the MD and SD of BP differences between 6 BPMs and Accutension Stetho. SBP1 and DBP1 are the unadjusted data. SBP2 and DBP2 were obtained from the biased blood pressure values measured by Accutension Stetho according to the deflation rate. BPM: blood pressure monitor.

2.6. Statistical analysis

The BPM‐Accutension differences for SBP and DBP were assessed separately. BP differences were calculated as the BPM reading minus the Accutension Stetho reading. SPSS 22.0 for Windows (SPSS Inc, Chicago, IL) and Microsoft Excel 2013 (Microsoft, Redmond, WA) were used for the analysis. Data were presented as the means ± SD. The correlation between the accuracy of the BP readings and the deflation rate was assessed using the Pearson correlation coefficient. The α‐level for a significant test was considered to be P < 0.05.

3. RESULTS

3.1. Performance of BPMs

Each BPM was in good condition. The reading agreements between the mercury sphygmomanometer and BPMs (readings displayed on the BPM screen) during the whole measurement process were consistent. All the BPMs worked properly with Accutension Stetho without affecting the accuracy.

3.2. Performance Of Accutension Stetho

The study was performed and analyzed from February 24, 2018, to April 19, 2018. Measurements of BP were successfully performed by the BPMs and Accutension Stetho. Each subject was measured by all 6 brands of BPM along with Accutension Stetho. Among one pair, 3 valid SBP and DBP values were obtained for each subject. Each BPM was measured in 99 subjects, and ultimately, 297 valid values were obtained from 99 subjects. There were no adverse events during the study.

3.3. Measured BP values

The percentages of SBP and DBP differences falling within different mm Hg (5, 10, and 15) are listed in Table 2. To avoid inappropriate interpretation of the data, viz., ranking the BPMs based on these data, we hid the brands and models of the studied BPMs in the table. After all, our intention is to introduce the clinical application of the Accutension Stetho, but not rank the BPMs. The same approach is used for other results.

The MDs and SDs for differences in SBP and DBP readings between different brands of BPMs and Accutension Stetho are listed in Table 3.

To eliminate the influence of the deflation rate, we used biased values to repeat the evaluation. An estimator for biased BP is P = P₀ ± (R/2‐1), where P₀ is the observed blood pressure and R is a random variable distributed over the cardiac cycle period. The details of correcting this bias can be found in the supporting information (Table S1).18

The deflation rates based on time (mm Hg/second) and beats (mm Hg/beat) are listed in Table 4 and were calculated using Formula 1 and Formula 2, respectively. An oscillometric BPM generally measures BP during slow deflation. When the machine finds the BP readings, it suddenly opens the valve to rapidly exhaust the air in the cuff (which means that the measurement is finished). The numbers of rapid exhaust that occurred prior to or close to the end of the Korotkoff sounds are also listed in Table 4. “Prior to” means the BPM rapidly exhausted prior to the end of the Korotkoff sounds. “Close to” means the gap between the rapid exhaust and the last Korotkoff sound was less than one heartbeat (Figure 3).

Table 4.

Number of readings with deflation prior to or close to the end of the Korotkoff sounds and the deflation rates of different BPMs

	Deflating prior	Deflating close	Deflation rate (mm Hg/beat)	Deflation rate (mm Hg/s)
A	0	11	2.23	2.44
B	0	6	3.48	4.04
C	1	1	6.10	6.93
D	4	10	2.44	2.79
E	8	6	3.66	4.22
F	1	9	4.85	5.48

The data list the number of readings prior to deflating and close to deflating (definitions can be found in the text).

BPM, blood pressure monitor.

Figure 3.

Examples of deflation problems. (A) One of the BPMs deflated close to the end of the Korotkoff sounds. The red arrow points to the deflating sound, which was near the last Korotkoff sound, and the gap between them was less than one heartbeat. (B) Another BPM deflated prior to the end of the Korotkoff sounds. The red arrow points to the deflating sound, which was followed by one Korotkoff sound

The correlation between the accuracy of the BP readings (percentages of SBP and DBP differences falling within 5 mm Hg were chosen as the main evaluation index) and the deflation rate was assessed using the Pearson correlation coefficient; they were not significantly correlated (r = −0.55 for SBP, P = 0.26; r = −0.45 for DBP, P = 0.37).

4. DISCUSSION

In this study, we designed a method to verify the readings of different electronic BPMs. Using a smartphone‐based app, doctors, as well as patients, could compare the BP values obtained from BPMs to those of Accutension Stetho. The Accutension Stetho values were viewed as a reference by using the auscultatory method to determine the BP values. The whole process could be videotaped and then played back, and the results are printable.

We also compared the accuracy among 6 different brands of BPMs. Similar to the ESH standard, we counted the number of BP differences between a given BPM and Accutension Stetho among certain mm Hg (5, 10 and 15). We then calculated the percentage of each mm Hg while using only that of 5 mm Hg to compare the accuracy of the BPMs. Model A had the highest accuracy of both SBP and DBP in this session, while monitors E and C had the worst accuracies in SBP and DBP, respectively. Similar to the AAMI standard, we also calculated the MDs and SDs of the BP differences between the 6 BPMs and Accutension Stetho. According to the results, monitors C and A had the best MD in SBP and DBP, respectively. Monitor A also had the best SD in both SBP and DBP. When the raw data were adjusted according to the deflation rate, monitor A had the best MD and SD in both SBP and DBP.

Basically, it is necessary to evaluate the accuracy of each electronic BPM before individual use. The greatest advantage of this app is that it verifies the BPMs simultaneously, which could minimize the sequential differences caused by traditional manual verification. Regarding how to determine the accuracy among different brands of BPMs, we mainly used the percentage of the BP difference (BPM‐Accutension Stetho) falling within 5 mm Hg, but we also took other factors into account: MDs, SDs of differences and the incidence of rapid exhaust prior to the end of the Korotkoff sounds; the latter of which is an important factor we introduced in this study. We noticed that some brands of BPMs rapid exhaust prior to or close to the end of the Korotkoff sounds, which made the DBP potentially high, so we evaluated the incidence. The definition of rapidly exhausting “prior to” the end of the Korotkoff sounds was that Korotkoff sounds could still be heard when the BPM started to deflate rapidly. Hence, the auscultatory reference for DBP was determined by the last Korotkoff sound available. Examples of rapidly exhausting “prior to” or “close to” the end of the Korotkoff sounds are illustrated in Figure 3. There was no occurrence of “prior to” with monitor A or B. All the BPMs in this study have “close to” phenomenon. Our study has shown that monitor A was ranked at the top, having the best performance for both SBP and DBP determinations here. In addition, the deflation rate of monitor A (the average was 2.23 mm Hg/beat) was within the range of 2‐3 mm Hg/beat, which was required by the auscultatory method. Did the deflation rate make monitor A so accurate, since we employed the auscultatory method to determine the difference? To eliminate the influence of the deflation rate, we used biased values to recalculate the score. It is well known that BP readings determined using the auscultatory method can be affected by the deflation rate, and studies have shown how to bias the value with a different approximate deflation rate. A method for correcting this bias is introduced in the supporting information.18 When we used the biased values to rank the 6 BPMs, the result remained unchanged. This finding is important because the same arm simultaneous method has been abandoned in the validation standard for theoretical concerns regarding the effect of fast deflation. However, if this effect could be eliminated by some correction, then the simultaneous method may not have been abandoned. According to the results, monitor A was ranked first among the 6 BPMs tested here. However, the results are limited to only the 6 BPMs included in this study and cannot be extended to other models of these brands. More models and brands will be assessed in future studies.

Although the MD/SD or percentage distribution of certain BPMs might have been acquired during the validation procedure, those subjects were enrolled differently, which made using those data to evaluate the accuracy impossible. However, we used the same subjects in our study, which made comparison possible. Another advantage of this study was that BP values were acquired simultaneously between the BPMs and Accutension Stetho. As we know, most of the BPM validations were performed in a sequential manner. In fact, if sequential readings are taken by a mercury sphygmomanometer and a test device, major inaccuracies can be detected.10 Usually, the first reading is higher than the following readings, so it is reasonable to discard the first reading, which is also recommended by the validation procedure. There is a tendency for BP to decline during this process. Therefore, it has been suggested that the subsequent readings be taken using a mercury sphygmomanometer and the test device alternatively.10 Unfortunately, the exact criteria for determining acceptability have not been established. In this study, we compared three readings taken simultaneously by 6 BPMs and Accutension Stetho. There existed a tendency for both SBP and DBP to decline from the first to the third reading, though there were no statistically significant differences (SBP: P = 0.59; DBP: P = 0.99; HR: P = 0.95) (see supporting Information).

The Accutension Stetho, as an inexpensive smartphone/stethoscope kit combination, was recommended to estimate precise BP values by auscultation to confirm the accuracy of an automated sphygmomanometer's readings in individual patient.17 Furthermore, with minimal training, most people can easily learn how to use the device independently following the instructions in the wizard.

There were some limitations to this study. First, some brands of BPMs, such as Panasonic and iHealth, determine the BP value in the inflation stage, which may cause too much noise to detect the Korotkoff sounds, so it is not suitable for Accutension Stetho to verify the accuracy among these brands. Second, Accutension Stetho does not interpret the auscultatory measurements on its own, though it could help both healthcare providers and patients check the accuracy of the BPMs they use. Additionally, we should not be too optimistic that an accurate measurement under one condition represents that the app can accurately measure the patient's blood pressure at all times. The accuracy could vary in the same individual when the BP values are different under different physiological conditions, which makes periodic validation necessary. Finally, our study was not intended to answer whether the deflation rate had an influence on the accuracy of the oscillometric method. However, although it seemed that a slow deflation rate contributed to the accuracy of the BPMs in this study, there was no statistical correlation between them. Further studies should be performed to test this hypothesis.

In a nutshell, although the exact origin of Korotkoff sounds is still not clearly known, the auscultatory technique remains the clinical gold standard for noninvasive blood pressure measurement. Using this technique, we demonstrated that a new, smartphone‐based tool named Accutension Stetho could help detect BP differences between the app and different electronic oscillometric sphygmomanometers. We were also able to approximately rank the accuracy of these BPMs. With this new tool, physicians and patients can verify the accuracy of certain BPMs easily and reliably during either diagnosis or long‐term blood pressure monitoring.

CONFLICT OF INTEREST

All the authors declare that they have no conflict of interest.

Supporting information

 

 

 

Zhang Z, Xi W, Wang B, Chu G, Wang F. A convenient method to verify the accuracy of oscillometric blood pressure monitors by the auscultatory method: A smartphone‐based app. J Clin Hypertens. 2019;21:173–180. 10.1111/jch.13460