Article Text

PDF

Original article
Development of a novel device for objective respiratory rate measurement in low-resource settings
  1. Hayley Turnbull1,
  2. Masumbuko Claude Kasereka2,
  3. Israel Amirav1,
  4. Sivasivugha Eugénie Sahika3,
  5. Ian Solomon4,
  6. Yossi Aldar4,
  7. Michael T Hawkes1
  1. 1 Department of Pediatrics, University of Alberta, Edmonton, Alberta, Canada
  2. 2 Université Catholique du Graben, Butembo, Congo
  3. 3 Association for Health Innovation in Africa, Butembo, Congo
  4. 4 RespiDx Ltd, Tel Aviv, Israel
  1. Correspondence to Dr Michael T Hawkes, Division of Infectious Diseases, Department of Pediatrics, University of Alberta, Edmonton, Alberta T6G 1C9, Canada; mthawkes{at}ualberta.ca

Abstract

Objective To evaluate a novel device (Respimometer) for objective measurement of respiratory rate (RR) in low-resource settings.

Design Description of prototype development, with proof-of-concept pilot field study at four paediatric healthcare facilities in Butembo, Democratic Republic of the Congo (DRC). The instrument was tested in healthy adult volunteers (n=10) and Congolese children (n=42) and compared with timed breaths (adults) or by reference comparator capnography (children). Correlation and Bland-Altman plots were generated for paired measurements.

Results The Respimometer is shaped like an oral thermometer and is placed in the mouth of the participants. RR is measured by thermistors positioned at the nasal outlet, which detect the temperature change between inhaled and exhaled breaths. In adult volunteers, the correlation coefficient between the delivered RR and the Respimometer measurement was median 0.992 (IQR 0.980–0.999). Measurement bias was −0.50 min−1 (95% CI −1.1 to +0.07, p=0.093), with upper and lower limits of agreement of −5.2 min−1 and 4.2 min−1, respectively. Among Congolese children, there was no evidence of bias: mean difference in RR +1.0 min−1 (95% CI −2.1 to +4.1, p=0.52). The upper and lower limits of agreement were −18 and +20 min−1, respectively.

Conclusion The Respimometer can accurately measure the RR in healthy adults and children in DRC. A simple and accurate instrument could facilitate the diagnosis of pneumonia by community health workers in low-income and middle-income countries, leading to reduced pneumonia-related deaths.

  • remote diagnostics
  • pneumonia
  • respiratory rate
  • child

Statistics from Altmetric.com

Introduction

Pneumonia is a leading cause of child mortality worldwide, causing close to a million deaths per year in children under 5 years of age.1 Countries in Africa and Asia report 2–10 times more pneumonia cases than industrialised countries and 70% of childhood pneumonia deaths occur in these low-resource communities.2–4 Pneumonia can be effectively treated with antibiotics, but only with accurate diagnosis and prompt treatment. The majority of pneumonia-related deaths occur in the community; only 25% of children with pneumonia in low-resource countries reach a healthcare facility.5 Therefore, in order to reduce childhood mortality from pneumonia, there is a need for community-based care. The WHO and Unicef recommend the integrated community case management (iCCM) model, delivered by community health workers (CHWs) to reach children with early diagnosis and treatment of pneumonia in low-resource settings (LRS).6

Tachypnoea is a sensitive clinical sign of lower respiratory tract involvement in a child with cough or difficulty breathing.7 8 In LRS, where there is scarce access to radiology and laboratory tests, the WHO and Unicef recommend that CHWs use the clinical presentation and respiratory rate (RR) for diagnoses and treat pneumonia in children according to specific case-management algorithms.6 CHWs are trained to manually count the RR by visually observing the child’s breathing, usually for 1 min. They then use the RR and to classify pneumonia severity and initiate antibiotic treatment according to the WHO’s Integrated Management of Childhood Illness (IMCI) strategy.9 Previous studies conducted in Africa have shown that RR assessed by CHWs is inaccurate.10 11 In a recent Ugandan study, CHWs counting the RR resulted in correct diagnosis and treatment of only 40% of all cases of childhood pneumonia.11 An instrument to facilitate accurate RR measurement at the village level could improve pneumonia diagnosis and case management.

Numerous technologies have been used to measure RR by detecting changes in parameters such as exhaled carbon dioxide, air temperature, humidity and chest wall movement (eg, capnography, fibre optic nasal sensors, nasal thermistors, nasal pressure transducers, piezoelectric monitor strips, pulse oximetry-derived plethysmograms, ECG-based RR derivation, respiratory inductance plethysmography and thermal imaging of expired air).12 13 Each method has strengths and limitations, and not all are suitable for LRS.14 15 To promote the development of pneumonia diagnostic aids that are appropriate for LRS, Unicef has developed a target product profile (TPP) which provides guidelines for the development of tools to help CHWs accurately measure RR in LRS.16 The TPP outlines suggested device performance and functional needs; for example, a device that can accurately measuring the RR in minimal time. Additionally, it provides information regarding user needs, suggested cost and context of use.

As an example of one product in current use, the Acute Respiratory Illness (ARI) timer is a field-ready handheld device that provides an audible time signal to assist counting breaths which has been used by CHWs in LRS.17 The device is simple, easy to use and does not rely on electricity. However, the device still relies on the CHW to count the RR, which has been found to be inaccurate. Additionally, smart phone apps and devices that detect chest wall movement, such as the Children’s Respiration Monitor (ChARM, Philips), have been developed.18–20

As a potential tool to aid in pneumonia diagnosis at the community level in LRS, we have developed a novel instrument, which we have named Respimometer. The device works by measuring the temperature waveform of the inhaled/exhaled air at the nasal outlet using sensitive thermistors. Here we present early data and proof-of-concept that the Respimometer is capable of accurate measurement of RR in healthy volunteers and Congolese children.

Methods

We followed an iterative procedure of prototype design and testing. Expert engineers (RespiDx) created an early version (Prototype 1), which was tested in the field among hospitalised children in the Democratic Republic of the Congo (DRC) to gather preliminary information about the accuracy, user-friendliness and acceptability of the device. Prototype 1 was tested in hospitalised Congolese children and the results were compared with a paediatrician counting and auscultating the RR as the reference standard. Based on data collected on RR measurements and qualitative user feedback from Prototype 1, engineers developed a second device (Prototype 2). This improved instrument was tested in healthy adult volunteers using timed RR to test the accuracy of the device under ideal conditions. Finally, Prototype 2 was tested in healthy and hospitalised children in the DRC, using capnography as the reference standard.

Setting

Field-testing of our device was conducted in Eastern DRC, a challenging environment marked by severe resource limitations in the health sector, intermittent violent conflict with population displacement and high pneumonia burden.21 22 The DRC has more pneumonia deaths than any other country except for India and Nigeria. Every year, approximately 148 000 children under five die of pneumonia in the DRC, accounting for 15% of child deaths in the country.5

Description of instrument

The Respimometer was shaped like an oral thermometer and made of waterproof medical-grade polymer. It was designed to measure the RR and other vital signs; here we present findings related to the RR measurement only. The device had a mouth piece, an upper lip flap incorporating two thermistors positioned at the nasal outlet (MF51E, Cantherm, Montreal, Canada) which detect the temperature-wave of exhaled air and a lower flap to stabilise the device in the mouth. Thermistor-based breathing sensors have been previous validated as a method of measuring RR.23 The output data of the sensors was transmitted by a cable to a laptop computer (Prototype 1) or a 10 cm × 10 cm stand-alone LED screen (Prototype 2) to display the readings. Figure 1 shows Prototype 2 used in a young Congolese child.

Figure 1

Photograph of Respimometer in a 4-year-old Congolese boy. The instrument is used like an oral thermometer for objective measurement of respiratory rate.

Instrument principle

Characteristics of the MF51E thermistor include small size, low thermal mass and rapid response to temperature changes. The device contained a microcontroller (CY8C54LP, Cypress Semiconductor, San Jose, California, USA) which includes a processor, memory and an analog-to digital conversion module. The output of the thermistors was converted to a digital stream of samples taken 20 times per second and filtered as follows: a high-pass filter removed high frequency ‘noise’, and a low-pass filter cleaned the data from longer-term temperature trends. The remaining digital data stream was then analysed to detect each peak and trough in the waveform. The duration of a breath was calculated as the time between consecutive peaks or consecutive troughs. The duration of each breath was then subjected to a ‘sanity test’ and counted only if it fell within a physiological range (10–150 breaths per minute). The counting of breaths started on detection of skin contact, when the thermistors are correctly positioned near the nares and continued for 30 s. The ‘peak to peak’ count and the ‘trough to trough’ count during 30 s were added together to calculate the breathing rate (breaths per minute). Waveforms from both nostrils were monitored, and measurements from the waveform with the stronger signal-to-noise ratio were used.

Testing Prototype 1 in hospitalised children

Prototype 1 was tested in hospitalised children in Butembo, DRC. The RR was simultaneously measured in two ways: (1) Respimometer and (2) paediatrician counting RR by observation and auscultation over 1 min. Observers for each measurement method were blind to the result of the other observer. We solicited qualitative feedback from end-users of the device (Congolese physicians).

Prototype 2 calibration in healthy adults

Healthy adult volunteers were recruited and coached to breathe at a fixed rate, according to a timer. The instrument was positioned in the mouth and volunteers shielded the device to ensure expired breaths were directed on the thermistors. Calibration curves were generated by recording the Respimometer reading across a range of artificially generated RRs.

Testing Prototype 2 in healthy and hospitalised Congolese children

Pilot studies in children were conducted in Butembo, DRC. We selected four health facilities: Centre Hospitalier Universitaire du Graben, Hôpital Général de Matanda, Centre Hospitalier Wanamahika and Poste de Santé La Guérison. Children were either admitted for inpatient care or were attending outpatient visits for vaccination or well-child monitoring. The RR measurements were taken by the study team and/or locally trained Congolese medical officers.

Statistical analysis

Among healthy volunteers, the RR measurements produced by the novel instrument were compared with timed breaths (reference comparator). Among children, the spontaneous RR was measured by Respimometer and, as reference comparators, paediatrician observation or capnography (Nellcor N-85 capnograph, Tyco Healthcare Group LP, Pleasanton, California, USA). One to three RR measurements were taken over approximately 5 min with both the capnograph and Respimometer and the mean result from each device was used for subsequent analysis. Correlation and Bland-Altman plots were generated for paired measurements. Data were analysed using GraphPad/PRISM software (V.5.0).

Ethics

The Comité d’Ethique du Nord Kivu (Université Catholique du Graben) and the Research Ethics Board of the University of Alberta approved the study. In the case of children, parents provided verbal informed consent to participate in the study, and for healthy adult volunteers, participants provided their own verbal consent.

Results

Prototype 1: hospitalised Congolese children

In an early field test, we used Prototype 1 in hospitalised Congolese children (3–4 December 2015). We included 23 children hospitalised at Matanda Hospital and Université Catholique du Graben, in Butembo, DRC. The median (range) age was 2.7 years (range 1 day to 9 years), and 15/23 (65%) were female. Admission diagnoses included respiratory tract infection, malaria, acute malnutrition and clinically suspected sepsis. Two children (8.7%) were uncooperative, could not have RR measured by the Respimometer and were excluded from further analysis.

The correlation and Bland-Altman plots for Prototype 1 are shown in figure 2. Correlation coefficient was r=0.44 (p=0.0497) comparing the RR using Prototype 1 to paediatrician gold standard measurement. Bland-Altman plot showed that the mean difference (bias) was −6.8 (95% CI −16 to 2.4) min–1, with upper and lower limits of agreement of −59 and +39 min–1, respectively. We noted that the waveform displayed on the laptop computer was sometimes irregular (noise), accounting for wide variability in the measurements. We also noted varying distance from the nares to the thermistors, depending on facial morphology and age of the child, which may have accounted for insensitive capture of temperature waveform in some children.

Figure 2

Early Prototype 1: Correlation and Bland-Altman plots comparing device to paediatrician observation among 23 hospitalised Congolese children. (A) Correlation coefficient was r=0.44 (p=0.0497). (B) Bland-Altman plot showed that the mean difference (bias) was −6.8 (95% CI −16 to 2.4) min–1, with upper and lower limits of agreement of −59 and +39 min–1, respectively. There was evidence of increasing error with increasing RR (p=0.0076).

Qualitative feedback was sought from users and patients’ parents. Users were positive about an instrument that could provide an objective quantification of RR. The device took a variable amount of time to display a result, which users felt was too slow. Positioning the thermistors at the upper lip/nose was not optimal for some patients and some patients feared the device. Users suggested that a nipple-shaped thermometer may be more acceptable to young patients. Prototype 1 elicited a gag reflex in some young infants, suggesting that the mouthpiece extended too far back into the oropharynx.

Based on the quantitative observations and qualitative feedback, an improved Prototype 2 was designed (figure 1). The mouthpiece was shortened, the position of the thermistors at the nasal outlet was optimised for younger infants and a compact LED display box (rather than a laptop computer) was created for display of the RR reading after a fixed 30 s interval.

Prototype 2: healthy adult volunteers

In order to define the performance characteristics of Prototype 2 under ideal circumstances, we initially tested Prototype 2 in cooperative adult volunteers who generated high volume, fixed rate respirations, using a timer. Ten healthy adult volunteers (nine male and one female) were recruited on the 29 and 30 June 2017, with median age 30 years (range 24–55). Prototype 2 accurately measured the RR from 10 to 65 min–1 (figure 3) in this population. In this range, the correlation coefficient between the delivered RR and the Respimometer measurement was median 0.992 (IQR 0.980–0.999). Bland-Altman plot (figure 3B) showed that the measurement bias was −0.50 min–1 (95% CI −1.1 to +0.07, p=0.093), with upper and lower limits of agreement of −5.2 min–1 and 4.2 min–1, respectively.

Figure 3

Improved Prototype 2: Accuracy and precision of RR measurement among 10 healthy adult volunteers. (A) Strong correlation was observed between standardised respiratory rate (timed breathing) and Prototype 2: median 0.992 (IQR 0.980–0.999). (B) Bland-Altman plot showed that the measurement bias was −0.50 min–1 (95% CI −1.1 to +0.07, p=0.093), with upper and lower limits of agreement of −5.2 min–1 and 4.2 min–1, respectively.

Improved Prototype 2: healthy and hospitalised Congolese children

Next, we examined the performance of Prototype 2 in children from the DRC. Forty-two children (48% female) were recruited between 23 and 27 June 2017, with median age 5 years (range 2 weeks to 12 years). The spontaneous resting RR was determined using capnography (reference comparator). Three children were unable to tolerate the nasal prongs for the capnograph and were excluded from further analysis. Correlation and Bland-Altman plots for Prototype 2 are shown in figure 4. There was no evidence of bias: mean difference in RR +1.0 min–1 (95% CI −2.1 to +4.1, p=0.52). The upper and lower limits of agreement were −18 and +20 min–1, respectively.

Figure 4

Improved Prototype 2: Performance in 42 healthy and hospitalised African children. (A) Correlation coefficient was r=0.47 (p=0.0099). (B) Bland-Altman plot showed no evidence of bias: mean difference +1.0 min−1 (95% CI −2.1 to +4.1, p=0.52). The upper and lower limits of agreement were −18 and +20 min−1, respectively. RR, respiratory rate.

Discussion

We have shown proof-of-concept that a novel instrument using thermistors positioned at the nasal outlet can accurately measure the RR in healthy adults and African children. The Respimometer (Prototype 2) provided unbiased measurement of RR in the range 10–65 min–1 in healthy adult volunteers. This range is adequate for the diagnosis of childhood pneumonia, since the threshold for tachypnoea, as defined for the purposes of iCCM is >50 min–1 in children 2–12 months of age and >40 min–1 in children 1–5 years of age.6 Beyond 65 min–1, the device need only display a ‘high’ readout, indicating tachypnoea and the need for antibiotic treatment. Previous studies have shown that CHWs sometimes have a difficult time interpreting numerical results and applying the RR appropriately to the ICMI treatment algorithm to determine appropriate candidates for antibiotic treatment.24 As such it has been recommended by the Unicef TPP that a device that provides automated results with a high level of treatment decision support is preferable.16 19 Additional alterations will be needed in order to improve device validity. Technological improvements to bring the thermistors into closer proximity to the nasal outlet are being pursued in order to maximise the signal to noise ratio at the thermistors.

The measurement of RR in children remains challenging. Non-contact devices like the ARI are thought to cause less distress to the child and therefore are less likely to alter the child’s RR. However, non-contact devices like the ARI and some smartphone applications still rely on the CHW to count the RR or tap the screen of the phone to determine the RR. Such methods are prone to error, leading to overdiagnosis and/or underdiagnosis and treatment. Other non-contact devices that are automated, such as infrared thermography and radar, are expensive high-technology devices that are not appropriate for use in LRS.12 13 19 RR measurement devices that come into contact with the child have been used; however, this is often noted to cause distress to the child which may alter the child’s RR. Studies in adult patients have shown mixed results on the effect of the use of airflow sensing methods on RR.25 Although oral thermometry is widely used in clinical practice, we found that some children were frightened by the Respimometer and/or unwilling to keep the device in their mouth. Nonetheless, this platform has potential to include an oral temperature sensor to detect fever, thus enabling simultaneous measurement of multiple vital signs. Additionally, design improvements to increase child-friendliness and acceptability are being explored.

Further issues will need to be addressed prior to the testing and implementation of the Respimometer at the community level. Some of the diagnostic aid attributes outlined by Unicef’s TPP include hygiene, charging, cost, portability and acceptability by CHWs and caregivers. In our study, we used 70% ethanol to sterilise the device between uses, rinsing with drinking water. For use in the community setting, availability of a cleaning method, instructions and supplies will be needed for CHWs. Access to clean water remains and issues for many communities. The device does require access to electricity. Additional developments will be needed to develop a device that requires minimal charging. The plan for the next prototype includes a battery-operated device with a display screen located on the device. Further data will be needed to outline device cost and operational life; however, based on cost of materials, the device may be deployed for approximately USD$20/device. Technical refinements and ‘ruggedisation’ of the device are required prior to field-readiness. For successful implementation at scale in the prehospital setting in LRS, mass production of a low-cost instrument will be necessary to equip the estimated 1.4 million first-line CHWs in the top 10 high pneumonia burden countries.16

Study limitations

This study has several limitations. We have provided early data and proof-of-concept that a thermistor-based instrument positioned at the nasal outlet can accurately measure RR; however, sensitivity and specificity relative to gold standard RR measurement in the key population of interest (children with pneumonia in the prehospital setting) and in the hands of target end-users (CHWs) remains to be demonstrated in community-based studies. The initial study population (median age 5 years) is older than the target population (children<5 years). For the Respimometer and reference comparator measurements, we used the mean RR from one to three measurements taken over 5 min. Additional studies will be needed to compare single measurements of the device to the reference comparator to reflect device use at the community level which will be a single measurement. Although further refinements and improved precision of the instrument are necessary prior to clinical use, this promising prototype could potentially fill the current need for a simple, low-cost instrument to objectively measure RR in LRS.

Conclusion

Here, we describe the development of a prototype for a simple device to objectively measure RR, with validation of instrument accuracy in healthy volunteers and African children. Although further refinement of the device is necessary prior to additional testing and clinical use, this initial prototype represents a promising first step towards objective RR measurement in children in a LRS. A device to objectively measure RR could augment CHW capacity for pneumonia diagnosis and management in low-income communities. Given the magnitude of paediatric pneumonia deaths (900 000 per year, globally) concentrated in prehospital settings in low-income and middle-income countries, improving the case management of pneumonia with appropriate technology has the potential for substantial impact on global child survival.

References

View Abstract

Footnotes

  • Contributors HT wrote and reviewed the manuscript. KMC performed the data collection, wrote and reviewed the manuscript. IA conceived the study, oversaw the data collection and reviewed the manuscript. SSE collected the data and reviewed the manuscript. IS designed and modified the device and reviewed the manuscript. YA designed and modified the device and reviewed the manuscript. MTH collected the data, performed the statistical analysis, wrote and reviewed the manuscript.

  • Funding This study was supported by USAID, the Association for Health Innovation in Africa (AFHIA) and Grand Challenges Canada.

  • Competing interests IS and YA are employees of RespiDx Ltd, a company which is developing the Respimometer for use in low-income settings as well as commercial markets in higher-income countries.

  • Patient consent Parental/guardian consent obtained.

  • Ethics approval Comité d’Ethique du Nord Kivu and Research Ethics Board of the University of Alberta.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Unpublished data from this study are available from the corresponding author on request.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.