The evidence for tapping out a respiratory rate
You are right to want this to be accurate. The evidence is there, and it is worth a look.
The short version
- Tapping took 22.8 seconds against 65.8 for a full minute, and was no less accurate. In adults.
- Against the 15 second shortcut most people actually use, tapping was better, and put more patients in the right early warning band.
- A 15 second count misses sick patients. Half the rates that should have scored 3 early warning points were missed by counting for 15 seconds.
- Rythmo itself has never been tested against a reference standard. Everything here is evidence for the method, not for my app.
I would rather you read the fourth point than the first three.
What I found when I reviewed all of it
I led a scoping review of respiratory rate measurement using mobile applications in healthcare settings, with colleagues from Victoria University and two Australian ambulance services. It mapped 14 studies, mostly cross-sectional and hospital based, and it is open access.
What we concluded:
- Most applications reported high usability and efficiency, although results of accuracy were mixed across the included literature.
- Median-based apps were more often an accurate measure of respiratory rate, though more research is required.
- Articles were commonly limited in generalisability by poorly defined reference standards, small sample sizes, or retrospective video recordings.
That last point is the honest state of play, and it is why this page is careful about what it claims. The method is promising. The evidence base is thinner than anyone would like.
Is it accurate if it is not a full minute?
The most common objection in the survey. It carries a hidden assumption worth pulling apart: that the full minute is what you are actually doing.
Mostly it is not. Seventy six per cent of the first 188 people who filled in the Rythmo survey count in their head against a watch, and the usual shortcut is 15 seconds multiplied by four.
That shortcut is the problem. On 770 readings from 321 acutely ill adult ward patients, a 15 second count sat 1.22 breaths per minute off a full minute, with limits of agreement from -7.2 to +4.7. More seriously, half the rates that scored 3 early warning points over a full minute were not identified by the 15 second count, and a quarter were missed by 30 seconds.
Tapping is not that shortcut. On 870 readings from 272 acutely ill adult medical patients, the tap algorithm was off by 0.11 breaths per minute where 15 seconds of observation was off by 1.85. It put 88 per cent of rates in the correct early warning band against 80 per cent for the 15 second count, and it took 16.2 seconds.
The studies, and the numbers
Adults, ward patients, tapping against the 15 second count. 870 respiratory rates on 272 acutely ill adult medical patients. Bias 1.85 breaths per minute for 15 seconds of observation and 0.11 for the algorithm, which took 16.2 seconds. Correct National Early Warning Score points in 88 per cent of algorithm rates against 80 per cent for 15 second observation. The authors concluded rates are measured nearly as quickly and more reliably by the algorithm. Nakitende I, et al. Acute Med. 2020;19(1):15-20.
Adults, shortcuts against the full minute. 770 rates on 321 patients. 15 seconds against a minute: bias -1.22, limits -7.16 to +4.72. 30 seconds against a minute: bias -0.46, limits -3.89 to +2.97. Rates scoring 3 points over a minute were missed by half the 15 second rates and a quarter of the 30 second rates. The authors concluded shortcuts often fail to identify sick patients. Rimbi M, et al. QJM. 2019;112(7):513-517.
Adults, tapping against the full minute. 59 nursing students measured adult patients from video. Tapping: bias 0.40, limits -2.86 to +3.67, mean 22.8 seconds. The full minute: bias 0.65, limits -2.11 to +3.41, mean 65.8 seconds. Suzuki R, et al. Jpn J Nurs Sci. 2022;19(3):e12481.
And the reference standard is not perfect either. A letter from the group behind the original method is titled, plainly, "Counting: an imprecise reference standard for respiratory rate measurement". Every limit of agreement on this page is measured against a human counting, which has its own error. Ansermino JM, et al. Pediatr Pulmonol. 2023;58(2):597-598.
Where it falls down
Tapping is quick and good on average. On a single reading in a sick child it can be a long way out.
The study that does not flatter the method
39 health care workers measured respiratory rate from videos of young children with acute lower respiratory infections, against a reference of two providers each counting for a minute. The tap app's bias was 0.4 breaths per minute, but its limits of agreement ran from -24 to +23 and its correlation with the reference was 0.62. Both apps tested agreed less well at higher respiratory rates. Spurr R, et al. Pediatr Pulmonol. 2022;57(12):3009-3016.
This is why Rythmo makes green cost more taps as the rate climbs. A fast rate has to earn its confidence.
The numbers behind the chart
| Study | Who, and what was compared | Mean | Limits of agreement |
|---|---|---|---|
| Suzuki 2022 | Adults on video, tapping | 0.40 | -2.86 to +3.67 |
| Suzuki 2022 | Adults on video, 60 second count | 0.65 | -2.11 to +3.41 |
| Gan 2015 | Children on video, tapping | 0.6 | -4.5 to +3.3 |
| Rimbi 2019 | Adult ward patients, 15 second count against a full minute | -1.22 | -7.16 to +4.72 |
| Spurr 2022 | Children with chest infections, tapping | -0.40 | -24 to +23 |
| Binotti 2019 | Newborn heart rate, simulation, tapping | 1 | -9 to +11 |
| Myrnerts Höök 2020 | Newborn heart rate, breathing, tapping | 3 | -14 to +19 |
| Myrnerts Höök 2020 | Newborn heart rate, not breathing, tapping | 3 | -48 to +53 |
Where else this matters
Newborn resuscitation
Newborn resuscitation runs on heart rate, usually auscultated for six seconds and multiplied by ten. This is the one setting where a randomised trial has tested what tapping changes, and it did not just change the number.
Tapping cut the time to starting chest compressions by 68 seconds and the first dose of adrenaline by 76 seconds.
Three studies, including the randomised trial
What it changes. A randomised controlled cross-over trial put 20 teams of three paediatric residents through a simulated newborn resuscitation, tapping against mental arithmetic. Time to the first heart rate call fell by 13 seconds, to starting chest compressions by 68 seconds, and to the first dose of adrenaline by 76 seconds. Cavallin F, et al. Arch Dis Child Fetal Neonatal Ed. 2020;105(1):41-44.
Accuracy. 160 heart rate measurements in high fidelity simulation agreed well with the set rate: kappa 0.80, mean difference 1 beat per minute, limits -9 to +11. Binotti M, et al. Arch Dis Child Fetal Neonatal Ed. 2019;104(4):F440-F442.
In the field. In Uganda, paediatricians counting breathing newborns against pulse oximetry were 3 beats per minute out on average, limits -14 to +19, in about 5 seconds. Midwives counting non-breathing newborns against ECG were also 3 beats out on average, but limits widened to -48 to +53. Myrnerts Höök S, et al. BMJ Paediatr Open. 2020;4(1):e000688.
Why respiratory rate at all
An abnormal respiratory rate predicts serious clinical events, and it is still the vital sign least likely to be written down.
The two papers worth knowing
Cretikos and colleagues made the case in 2008 and called it the neglected vital sign. Documentation of vital signs in many hospitals is extremely poor and respiratory rate in particular often is not recorded, despite substantial evidence that an abnormal rate predicts potentially serious events. Med J Aust. 2008;188(11):657-659.
In 2017 a group looked at recorded rates from 36,966 hospital admissions. They clustered on 18 and 20 per minute instead of spreading out the way machine-measured heart rate does. The authors concluded the rate was being estimated rather than counted. BMJ Qual Saf. 2017;26(10):832-836.
How to get a good reading
From the practical guidance in my review for Australian Paramedic.
- Use it when you would otherwise estimate or guess, when there is no time for a 30 second count, or when capnography is not available.
- Do not use it for agonal breathing, extremely irregular patterns such as Cheyne-Stokes, or anywhere needing continuous monitoring.
- Get a good reading by measuring while the patient is not talking, watching the abdomen rather than the chest wall, tapping on every inhalation, and taking at least five taps, more if breathing is fast or irregular.
Read the practical review in Australian Paramedic, spring 2025
What the evidence changed in Rythmo
Every design decision below exists because of something in the literature above.
- Median gap between taps, not the average
- Published method. One mistimed tap cannot drag the number.
- Rejects a gap more than 13 per cent from the running median
- Published method. Excluding inconsistent taps improved accuracy.
- Red until 3 valid taps, green not before 5
- Accuracy improves with more intervals, so a three-tap count should not look confident.
- Green costs more taps as the rate climbs
- Agreement gets worse at higher rates.
- Irregular flag from 10 taps, held at amber
- A median is a poor summary of an irregular pattern.
- Fresh baseline after 3 rejected taps
- My addition, not from the published method.
Try it yourself in the free Rythmo apps, or tap along with the demo on the home page.
What Rythmo has not shown
- No Rythmo app has been tested against a reference standard. Not against a capnograph, not against an ECG, not against an experienced clinician counting for a minute.
- Three parts of the engine are mine and unpublished. The fresh baseline, the rising green threshold, and holding at amber while Irregular is flagged.
- The handheld device does not exist. Every image of it is a concept render.
- Nobody has measured the distraction risk. Survey responses raised it, that tapping might pull your attention off the chest. I looked for a study and could not find one. The method removes the arithmetic, not the observation. You still have to watch every breath.
If you work somewhere this could be tested properly, I would like to hear from you at hello@rythmo.com.au.
References
Every source on this page
- Sallabank L, Oswald J, Willett S, Kelleher J, Haskins B. Respiratory rate measurement using mobile applications in healthcare settings: a scoping review. Healthc Technol Lett. 2026;13(1):e70035. doi.org/10.1049/htl2.70035
- Sallabank L. Tap-per-breath apps. Australian Paramedic. Spring 2025:12-15.
- Nakitende I, Namujwiga T, Dunsmuir D, Ansermino JM, Wasingya-Kasereka L, Kellett J. Respiratory rates observed over 15 seconds compared with rates measured using the RRate app. Acute Med. 2020;19(1):15-20.
- Rimbi M, Dunsmuir D, Ansermino JM, Nakitende I, Namujwiga T, Kellett J. Respiratory rates observed over 15 and 30 s compared with rates measured over 60 s. QJM. 2019;112(7):513-517. doi.org/10.1093/qjmed/hcz065
- Suzuki R, Takada T, Takeshima T, et al. Usefulness of a mobile phone application for respiratory rate measurement in adult patients. Jpn J Nurs Sci. 2022;19(3):e12481. doi.org/10.1111/jjns.12481
- Ansermino JM, Ginsburg AS, Dunsmuir D, et al. Counting: an imprecise reference standard for respiratory rate measurement. Pediatr Pulmonol. 2023;58(2):597-598. doi.org/10.1002/ppul.26208
- Karlen W, Gan H, Chiu M, Dunsmuir D, Zhou G, Dumont GA, Ansermino JM. Improving the accuracy and efficiency of respiratory rate measurements in children using mobile devices. PLoS One. 2014;9(6):e99266. doi.org/10.1371/journal.pone.0099266
- Gan H, Karlen W, Dunsmuir D, Zhou G, Chiu M, Dumont GA, Ansermino JM. The performance of a mobile phone respiratory rate counter compared to the WHO ARI timer. J Healthc Eng. 2015;6(4):691-703. doi.org/10.1260/2040-2295.6.4.691
- Spurr R, Ng E, Onchiri FM, et al. Performance and usability of a new mobile application for measuring respiratory rate in young children with acute lower respiratory infections. Pediatr Pulmonol. 2022;57(12):3009-3016. doi.org/10.1002/ppul.26125
- Binotti M, Cavallin F, Ingrassia PL, Pejovic NJ, Monzani A, Genoni G, Trevisanuto D. Heart rate assessment using NeoTapAdvancedSupport: a simulation study. Arch Dis Child Fetal Neonatal Ed. 2019;104(4):F440-F442. doi.org/10.1136/archdischild-2018-315408
- Cavallin F, Binotti M, Ingrassia PL, et al. Impact of a mobile application for heart rate assessment in simulated neonatal resuscitation: a randomised controlled cross-over study. Arch Dis Child Fetal Neonatal Ed. 2020;105(1):41-44. doi.org/10.1136/archdischild-2018-316757
- Myrnerts Höök S, Pejovic NJ, Cavallin F, et al. Smartphone app for neonatal heart rate assessment: an observational study. BMJ Paediatr Open. 2020;4(1):e000688. doi.org/10.1136/bmjpo-2020-000688
- Cretikos MA, Bellomo R, Hillman K, Chen J, Finfer S, Flabouris A. Respiratory rate: the neglected vital sign. Med J Aust. 2008;188(11):657-659. doi.org/10.5694/j.1326-5377.2008.tb01825.x
- Badawy J, Nguyen OK, Clark C, Halm EA, Makam AN. Is everyone really breathing 20 times a minute? BMJ Qual Saf. 2017;26(10):832-836. doi.org/10.1136/bmjqs-2017-006671
Disclaimer
Rythmo is a reference and training tool. It reports a rate calculated from your own taps and displays published reference ranges for comparison. It does not diagnose, does not decide treatment, and does not replace your clinical assessment. Confirm every reading before you act on it or document it.