Validity of spirometry performed online

RESULTS

283 patients were initially selected. 22 (7.8%) patients were excluded due to lack of written consent and psychophysical incapacity (two patients were unable to understand the instructions and one patient had thoracic pain). The remaining 261 were randomised. In two patients (one in the conventional and the other in the online starting group), it was impossible to obtain any numerical spirometric values, because of nausea in one case and absence of cooperation in the other. In addition, a patient starting in the online group could not introduce the filter into the spirometer handpiece, so the online spirometry could not be performed. These three patients were analysed as nonacceptable spirometries (two in the online group and one in the conventional group) in the intention to treat analysis.

Excluded patients were similar to included subjects with regards to sex, body mass index, cigarettes smoked per day, arterial oxygen saturation (S_a,O2) and reason for consultation. They were statistically different in age (mean±sd 59±17 yrs in excluded versus 49±14 yrs in enrolled; p = 0.01).

Table 1 shows patient characteristics. Subjects were middle-aged, predominantly male and overweight. Most of patients were smokers or ex-smokers with some degree of dyspnoea (generally mild). Obstructive patterns (based in conventional spirometry) were present in 16.6% of the sample and nonspecific pattern in 5.9%, according to predicted values 32.

No significant differences were found between online and conventional spirometric values in either the intention to treat or per protocol analyses, with narrow 95% confidence intervals of the mean difference (table 2). Figure 1 shows the agreement of FVC and FEV₁ between online and conventional methods using the Bland–Altman test. Although the biases between the two methods were small, the limits of agreement were higher in the intention to treat analysis (see Discussion for details). The difference between online and conventional spirometry was independent of measurement size.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1–

Bland–Altman analysis for a, c) forced vital capacity (FVC) and b, d) forced expiratory volume in 1 s (FEV₁) performed online and using conventional methods in the a, b) intention to treat and c, d) per protocol analyses. ······: mean difference; ---: limits of agreement.

In the intention to treat group, there were 54 (21%) patients off-range in FVC and 40 (15%) off-range in FEV₁. In the per protocol group, there were 29 (14%) patients off-range in FVC and 21 (10%) off-range in FEV₁. We did not find statistical differences in the comparisons between off-range and nonoff-range groups in the variables from table 1 in either the intention to treat or per protocol groups of patients. In the off-range group, the agreement for the determination of nonspecific and obstructive patterns between conventional and online spirometries was as follows: 1) in the intention to treat group, the κ-test results were 64% (p<0.001) for the nonspecific pattern and 77% (p<0.001) for the obstructive pattern; and 2) in the per protocol group, the κ-test results for the nonspecific pattern were 79% (p<0.001) and for obstructive pattern was 86% (p<0.001).

Both acceptability (conventional 88.9% and online 82%; p = 0.001) and acceptability/repeatability (conventional 87.7% and on-line 81.6%; p = 0.005) comparisons were statistically different (∼6% better in the conventional procedure). The number of spirometric manoeuvres needed to achieve acceptability and repeatability criteria was marginally higher in the online method and this type of spirometry procedure tended to take longer time (table 2). No association was found between age and the differences in spirometric values, number of manoeuvres and time spent between the two spirometry methods.

The starting procedure had a weak influence on the percentage of patients with spirometric quality criteria (acceptability and repeatability). There were statistically significant differences between the two procedures in the online but not the conventional starting groups, showing a learning effect (table 3). However, the comparison between the two starting groups in the percentage change (difference) between the two spirometric methods was statistically significant in acceptability but did not achieve significance in acceptability and repeatability, indicating a limited learning effect. In addition, the difference between conventional and online when both procedures were the first to start was only 5.9% (85.5% for conventional and 79.6% for online) (table 3).

DISCUSSION

This large, controlled, randomised study is the first published study to ascertain the validity of high-quality spirometry performed online. The spirometric values achieved online were very similar to those obtained using the conventional procedure. The reference (conventional) had a higher percentage of patients with spirometric quality criteria and a lower number of manoeuvres to achieve them.

From a practical point of view, the importance of the percentage of patients with spirometric quality criteria (acceptability and repeatability) is shown by the difference between conventional and online results when both procedures were the first to start, given that this situation reproduces normal conditions, that is, 85.5% for conventional and 79.6% for online, similar to the values obtained when analysing the total sample. In addition, there was a slight learning effect. In summary, both result in better quality criteria in the conventional spirometry. However, a difference of 5.9% is not crucial because variation in quality criteria between different respiratory function laboratories, or even between different technicians in conventional spirometry, exceeds this percentage 25, 33–35. Conversely, although a greater number of spirometric manoeuvres and more time were needed for the online procedure in comparison with the conventional procedure, the mean differences were 0.5 additional manoeuvres and 0.7 min, which may be acceptable as well.

The spirometry is intrinsically variable between different tests in the same time period and patient: values ≤150 mL in both FVC and FEV₁ are not considered significant 14. When we calculate the agreement limits of the differences between the two methods in the Bland–Altman plot, we include this variability in addition to the variability between both types of spirometries. Since intrinsic variability is not systematic (it can occur in both methods of spirometry and vary among patients), the mean difference between conventional and online spirometries may be close to zero. For this reason, agreement limits can be overestimated. Based on the 95% confidence interval of the mean difference between methods, the probability of finding wide differences is low (table 2). In any case, agreement limits were higher in the intention to treat analysis. From a practical point of view, the most important analysis is per protocol, because spirometries without quality criteria are not taken into account.

In order to better characterise the discrepancies between the two spirometric methods, we identified off-range patients as having differences in FVC and FEV₁ greater or less than 150 mL (repeatability criteria) between online and conventional spirometries. Although we did not find predictive variables for this off-range group, the frequency of off-range scores was low in the per protocol group (14% in FVC and 10% in FEV₁). This frequency was very similar in the 54 patients in whom conventional spirometry was repeated twice for inter- and intra-observed agreement (12% in FVC and 9% in FEV₁), meaning that most of the numerical differences between online and conventional spirometry can be explained by the spirometric technique itself. In addition, high agreement between online and conventional spirometries was found to determine nonspecific and obstructive patterns in the off-range group.

New information and communication technologies can change many procedures that are carried out in daily clinical practice. Previous studies have shown that experienced patients can carry out spirometry at home to monitor FEV₁ in asthma and the respiratory state of lung transplant recipients, transmitting data by telemetry 27, 28. Moreover, telemetry supervision of spirometry quality by specialists could be a teaching method for improving the quality of spirometry in primary care 13, 36. The main subject of our study is the potential to relocate conventional spirometry to any place where a high-quality spirometer could be used efficiently, but its precise role in the primary care landscape has yet to be defined.

Proper technical training for performing spirometry is essential to achieving adequate quality. While the recommended spirometry training time for pulmonary function laboratory technicians has been established at 6 months 37, there is no recommendation for primary care staff. The time employed in different studies is shorter and highly variable 12, 25, 26. In addition, to provide adequate initial training and periodic in-service training for a large staff (e.g. all primary care centres) is expensive and logistically complex. As we understand it, the best option for large primary care centres, with many patients needing spirometric studies, may be to perform the spirometries in a place with proper and periodic training, perhaps by telematic learning 8, 13, 36, 38. However, in smaller primary care centres with fewer patients and resources to professionally train workers to carry out spirometry, the online alternative may be preferable. Similar reasoning could be used for small hospitals without pulmonary function laboratories.

Our randomised sample resulted in self-selection for relatively young patients. The main reason was the lack of written consent. The increased time needed to undergo an unknown test may be an important inconvenience for older patients. Nevertheless, this bias does not seem to influence our results, since age was not associated with the difference between both spirometric procedures in primary and secondary outcomes. On the other hand, most studied patients did not have altered spirometries because we selected the patients in order to obtain a representative sample of spirometrically naïve patients from the population referred by general practitioners for pulmonary consultation. In any case, the subgroup of patients with altered spirometries showed no statistical differences between numerical variables (FVC, FEV₁ and FEV₁/FVC) registered with either spirometric method. Thus, the results should be similar in a sample with more severe spirometric alteration.

We included patients with suspected sleep apnoea to prevent the selection of patients referred from primary care for pulmonary consultation and because a spirometry at a hospital or primary care centre is now normally indicated in the diagnosis of these patients. In this group, 60% were smokers or ex-smokers, 39% had some grade of dyspnoea and 17% had chronic cough. Therefore, many of these patients could be referred for spirometric testing for reasons other than suspected sleep apnoea.

Online spirometry should be used via internet or intranet. Both, but particularly the internet, have important variability in the bandwidth available at any given point in time. Our required bandwidth was very low: 512 Kbps of symmetric bandwidth (for upload and download traffic). It is easy to achieve a guaranteed minimum symmetric bandwidth of 512 Kbps from internet or intranet providers over current infrastructure.

As already mentioned, once the patient entered the spirometer room, they had no technician present to help with proper placement of the bacterial filter, for holding the spirometer hand piece or for starting the process. Despite this apparent difficulty, only one patient was unable to put the bacterial filter in place. We believe that improving the written and graphical explanation provided before the test could improve patient cooperation and adjustment performance.

Light-headedness or dizziness are frequent during a spirometry procedure and these occurred during the telematic approach. All subjects were assisted telematically by technicians. Although during the study (624 complete spirometric procedures in 258 patients) we did not observe other important adverse problems, such as pneumothorax, bronchospasm or fainting, they obviously can occur. The latter has a special relevance because subjects could fall down and injure themselves. In our study, the technician was positioned close to the patients and they could respond rapidly, but in real practice the technician should be located far from the subject. For this reason, we understand that online spirometry requires some face-to-face assistance by a trained person in the spirometry room (see later).

We tried to replicate the conditions found outside the pulmonary function laboratory, but the simulation was not complete. Therefore, the following issues must be taken into account for the online spirometry to be judged practical. 1) Patients need attention before coming to the spirometer room in order to understand where and when the spirometry should be done, and must receive written information; an assistant and/or audio/video support should be available. 2) Management of the spirometer and its accessories and daily calibration check require trained staff. 3) Assistance for fainting and other potential adverse problems requires trained staff. 4) The application of an inhaled bronchodilator for assessing bronchodilation also requires trained staff. In summary, online spirometry cannot be done without face-to-face support, but the level of training and staff qualification can be much lower for online than for conventional spirometry. In any case, adequate quality control procedures 39 should be implemented for online spirometry.

A limitation of our study is that the economic cost was not analysed. Nevertheless, it can be inferred from the additional equipment required: a computer with a basic webcam for teleconferencing in the spirometer room, two computers (one with a basic webcam) for remote control of spirometer software and teleconferencing in the technician's room (respiratory function lab) and an intra- or internet connection. Regarding staff costs, the previously mentioned resources spent on direct, face-to-face support by the hospital technician will be saved. Due to the fact that the training and salary of a hospital technician should be higher than that of primary care staff providing face-to-face assistance, additional primary care staff should not result in additional costs, at least in theory. In comparison with conventional spirometry performed in primary care by very trained staff (technicians), the salary of hospital and primary care technicians in our public health system is similar, but in other systems, hospital technicians’ salaries can be higher, which should be added to the total cost of the online procedure. Nevertheless, the additional cost does not seem too high, considering that the proposed procedure could solve the problem of performing technically acceptable spirometry in primary care.

In conclusion, spirometry performed online can be a good alternative to conventional spirometry. Its use could be reserved for intermediate and small primary care centres and hospitals without a pulmonary function laboratory.