A service evaluation of the accuracy of electronic prescriptions used to calculate nebulised medication adherence in adult with cystic fibrosis

Introduction

Adherence to nebulised medications in people with cystic fibrosis (CF) is known to be less than 50% (Daniels et al. 2011; Quittner et al. 2014). Suboptimal adherence leads to increased pulmonary exacerbations and CF-related hospitalisations (Eakin et al. 2011; Quittner et al. 2014). Monitoring adherence to nebulised medications is necessary for ‘effective and efficient treatment planning’ (Sabaté 2003). It enables clinicians to determine whether a poor response to treatment is genuine, requiring a change in medication and a possible increase in treatment costs or treatment burden, or whether the poor response is due to suboptimal adherence.

Accurately measuring medication adherence can be challenging, particularly for people with CF whose nebulised medication regimens can be complex (Sawicki et al. 2013). For example, nebulised antibiotics are often prescribed on an alternate month basis for people with chronic pseudomonas aeruginosa infection. During the ‘month off’ a different nebulised antibiotic may be prescribed or no antibiotic at all (NHS England 2014). Some nebulised antibiotic medications are prescribed twice a day whilst others are prescribed three times a day; nebulised medications may also be stopped for a few days if haemoptysis occurs (Cystic Fibrosis Trust 2017).

Electronic prescriptions (e-prescriptions) can be used to provide a denominator from which medication adherence can be calculated. This objective measure is not subject to recall or report biases, unlike adherence measures more commonly used in clinical practice such as patient recall, and can provide real-time data (Forbes et al. 2018). Electronic measuring devices, such as the iNeb (Philips Respironics, Chichester, U.K.) or eTrack (Pari GmBH, Germany), which record when a nebuliser device is used to take a nebulised medication, provide objective data that can be compared to the e-prescription to calculate medication adherence.

The CFHealthHub data observatory study is a multi-centre study measuring nebulised medication adherence, in people with CF, using the iNeb and eTrack chipped nebuliser devices. The devices record each time a nebuliser is completed and compare this to the total doses on the e-prescription, calculating an adherence percentage of the patients’ daily target. At the Wessex Adult CF Centre, the e-prescription is entered into the CFHealthHub website by the research interventionist and must be updated when any changes are made to the nebulised medication prescription by the CF team. Therefore, inaccurate e-prescriptions will affect the accuracy of adherence calculations. Clinicians and patients use the adherence calculations to assess nebulised medication efficacy. Therefore, it is important that e-prescriptions in CFHealthHub remain accurate in order to accurately calculate medication adherence and provide clinicians with a useful tool to guide decision making.

Prescription complexity is known to increase the inaccuracy of e-prescriptions (Ryan et al. 2014). However, to date the accuracy of CFHealthHub e-prescriptions remains unknown. This service evaluation sought to review the accuracy of e-prescriptions in CFHealthHub at the Wessex Adult CF Service, where the author had access to the e-prescriptions of the participants. It aimed to understand the nature and causes of any inaccuracies found and, furthermore, to assess the effect of prescription complexity on prescription inaccuracy as well as types of error.

Methods

Eligibility

Participants were included in the service evaluation if they were using a chipped nebuliser device as part of the CFHealthHub study and they either attended an outpatient clinic appointment or were admitted for inpatient treatment between March and May 2018 inclusively. Participants were excluded from the service evaluation if they were unable to stay to discuss their two prescriptions at the end of their clinic appointment, or if they were unavailable on at least two separate occasions during their inpatient stay.

Procedure

Usual procedure for any participant enrolled in the CFHealthHub study was for the research interventionist to check the CFHealthHub e-prescription with the participant during a clinic visit or whilst they were receiving inpatient care. The e-prescription was checked against documentation in the electronic health record (EHR), for example, in clinic letters or home delivery prescriptions, as it is considered by the Wessex Adult CF Service to be the most accurate and up-to-date record of a patient’s current prescription. Therefore, in the service evaluation the EHR record was considered the ‘gold standard’ prescription. The CFHealthHub e-prescription was identified from the CFHealthHub online platform. Figure 1 shows the process for comparing the two prescriptions.

See Figure 1: Flow diagram demonstrating the protocol used to compare gold standard and CFHealthHub prescriptions during the service evaluation.

Participant and prescription characteristics were also captured from the EHR for each participant in the service evaluation. All data were stored in password protected electronic files.

In line with guidance from the Health Research Authority and from the research and development department at University Hospital Southampton, this was not considered to be a research study since participants in the service evaluation continued to receive usual care, were not randomised to different groups and the project did not seek to generalise results. Therefore, ethics and approvals were not required. However, the service evaluation was approved and registered at University Hospital Southampton (SEV/0066). The CFHealthHub data observatory study has received ethics and approvals from the London-Brent Research Ethics Committee (17/LO/0032).

Data analysis

A mixed methods approach was used to analyse data from the service evaluation. Quantitative data were used to identify where e-prescription inaccuracies existed and to identify any associations between e-prescription inaccuracies and prescription complexities. Qualitative data were then used to further understand how the inaccuracies may have occurred.

Quantitative data analysis was carried out to analyse the following:

1. The proportion of CFHealthHub prescriptions containing an inaccuracy when compared to the ‘gold standard’ prescription.

2. The types of prescription error found.

3. The association between prescription accuracy/prescription error types and prescription complexity, as defined in Table 1.

Table 1: Prescription complexity variables analysed in the service evaluation.

Due to the categorical nature of the data and small sample size, Fisher’s Exact test was used to assess the association between prescription accuracy and prescription variables. Additionally, the difference in proportions of inaccurate and accurate prescriptions for different prescription variables was analysed to demonstrate the size of any observed associations.

A qualitative approach, using thematic analysis (Braun & Clarke 2006), was used to analyse written summaries of discussions with participants to elucidate themes relating to the causes of prescription inaccuracies. Braun and Clarke (2006) use a reflexive approach to thematic analysis and this was chosen for its flexibility, including the ability to use it with many different types of data, including summaries of discussions.

Results

A total of 35 eligible participants were identified. However, three were excluded as they were unable to stay to discuss their prescriptions at the end of their clinic appointment and two participants were unavailable on the ward to discuss their prescription during their inpatient stay. Characteristics of the 30 participants included in the service evaluation and their prescriptions have been summarised (Table 2).

Table 2: Demographics of participants and their prescriptions included
in the service evaluation.

IQR = interquartile range.

1. Prescription accuracy

A total of 13 (43%) CFHealthHub e-prescriptions were found to contain at least one error leading to inaccuracy when compared to the gold standard prescription. Four CFHealthHub e-prescriptions contained two errors and one e-prescription contained four errors. In total, this caused adherence to be underestimated for five participants and overestimated for four participants, whilst two prescriptions contained on e-prescription so no adherence calculation could be made. Further analysis showed a significant association between prescription inaccuracy and prescriptions containing an alternating medication regimen (p = 0.025), although there is a wide confidence interval for the differences in proportions and a relatively small sample size, Table 3.

Table 3: Association between CFHealthHub inaccuracies and prescription complexities.

*Significant at <0.05.

2. Prescription error types

There were four types of prescription error found (Table 4). Medication list errors were shown to be significantly associated with prescriptions containing an alternating medication regimen (p = 0.007), Table 5.

Table 4: Errors identified in CFHealthHub e-prescriptions.

Table 5: Association between ‘medication list’ errors and prescription complexities.

*Significant at <0.05.

3. Causes of prescription inaccuracy

There were three causes of prescription error found through analysis of the EHR and thematic analysis of discussions with participants about their prescriptions (Figure 2). Prescription changes were the most frequent cause of prescription inaccuracy. This occurred when a change was made to the gold standard prescription but the CFHealthHub e-prescription was not updated; for example, the prescription may have been changed following an outpatient appointment.

‘…found that dornase made him tight chested so it was agreed at his last clinic appointment… that he should alternate between dornase one month and hypertonic saline one month’ – participant 24.

The second cause of prescription inaccuracy occurred when the e-prescription was entered incorrectly. Errors could persist for several months.

‘… reports that… she stopped Colomycin just before May 2018’ – participant 1.

The final cause of prescription inaccuracy was found to be linked to the use of two different nebuliser devices by the same participant. Typically, one device was used to take one or two nebulised medications and the other device was used for other nebulised medications. However, only one of the devices was linked to the participants’ CFHealthHub account. Errors occurred when the participant switched which device they used to take their different medications but did not alert the clinical team.

‘… iNeb was broken recently so she took her Dornase via her eTrack during April/May’ – participant 26.

Discussion and conclusion

This service evaluation has highlighted that 43% (13/30) e-prescriptions in an online platform, CFHealthHub, measuring adherence to nebulised medications, were inaccurate. Consequently, nine participants’ nebulised medication adherence was overestimated or underestimated. Although the service evaluation did not seek to determine the effect of inaccurate e-prescriptions, it does highlight the need to maintain an accurate e-prescription, particularly when using adherence data to inform treatment effectiveness and MDT decision making.

Evidence of e-prescription inaccuracy rates varies considerably in the literature from <1% to >80%. This is largely due to the different criteria used to define a prescription error making it difficult to compare the error rate found in this service evaluation with those observed in other studies (Jayawardena et al. 2007; Velo & Minuz 2009; Kaushal et al. 2010). Studies often include ‘prescribing errors’, that is errors that occur when making a clinical decision about a prescription but CFHealthHub e-prescriptions are not used to dispense medications and therefore do not contain prescribing errors.

Only one prescription complexity, alternating medication regimen, was found to be significantly associated with prescription inaccuracy. Unlike medications prescribed on a continuous basis which can be entered into the e-prescription once, alternating medications must be entered on each alternate month, increasing the chance for an error to occur. More frequent quality checks and/or an alert system may help address this. Although there was not a statistically significant association between prescription inaccuracy and prescriptions containing >2 medications, there was a 30% difference in the proportion of inaccurate prescriptions with >2 medications. This suggests a trend towards e-prescription inaccuracy with a greater number of medications in the prescription. Clinicians at the Wessex Adult CF service should consider whether e-prescriptions with alternating regimens or >2 medications should be checked on a more frequent basis to improve accuracy.

There were three causes of prescription inaccuracy found. However, due to the small sample size it is unlikely data saturation was reached and further causes of e-prescription inaccuracy may exist. Inaccuracies caused by a failure to update the e-prescription after a change was made to the gold standard prescription suggest communication deficiencies within the CF team that need to be addressed.

Although no statistically significant association was found between e-prescription inaccuracy and using two different nebuliser devices, the use of two devices emerged as one of the themes leading to prescription inaccuracy. Since only four participants (13%) were found to be using two different nebuliser devices and given the small sample size in the service evaluation there is an increased risk of a type 2 error, which may explain these apparently conflicting results. Further analysis with a larger sample size is needed to understand the effect of using two different nebuliser devices on e-prescription accuracy.

This service evaluation has several limitations. Firstly, although 43% of e-prescriptions in the service evaluation contained an alternating regimen, only 22% of all CFHealthHub e-prescriptions at the Wessex Adult CF Service contain an alternating prescription. Therefore, it may have overestimated the percentage of inaccurate e-prescriptions. Secondly, the service evaluation is context-specific and may have limited generalisability to other CF centres.

Strengths of the service evaluation include the use of a mixed-methods design which allowed the service evaluation to reveal the issues in greater depth than a purely qualitative or quantitative approach would have allowed. Finally, this service evaluation is the first within the Wessex Adult CF Service to look at the accuracy of e-prescriptions which are used to measure nebulised medication adherence in people with CF. It highlights the wider challenges of measuring adherence in this patient group, the types of prescription that are more prone to inaccuracy and suggests areas for improving e-prescriptions accuracy.

In conclusion, this service evaluation has underlined the difficulties of maintaining accurate e-prescriptions for people with CF, and it has highlighted the causes of e-prescription inaccuracy at the Wessex Adult CF Service, although the full extent of e-prescription inaccuracy may not have been identified in this sample size. There is a need to introduce strategies to improve the accuracy of these e-prescriptions to ensure that the adherence data obtained from them remains reliable and can be used to optimise patient care.

Key points

1 Measuring adherence to nebulised medications requires attention to the prescription to ensure adherence calculations are accurate. This is particularly challenging in CF due to prescription complexity.

2 Prescriptions that contain an alternating medication regimen and more than two medications may be at increased risk of prescription inaccuracy.

3 Effective communication between different members of the CF MDT is key to ensuring e-prescription accuracy when changes are made to the nebulised medication prescription.

Acknowledgements

Thank you to the staff and patients at the Wessex Adult CF Service, University Hospital Southampton.

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Ethical and R&D approval

This service evaluation did not require ethical approval and was approved and supported by the clinical and management teams.

Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101. https://www.tandfonline.com/doi/citedby/10.1191/
1478088706qp063oa.

Cystic Fibrosis Trust. (2017). Standards of care and good clinical practice for the physiotherapy management of cystic fibrosis. https://www.cysticfibrosis.org.uk/sites/default/files/2020-12/Standards%20of%20Care%20and%20Good%20Clinical%20Practice%20for%20the%20Physiotherapy%20Management%20of%20Cystic%20Fibrosis%20Fourth%20edition%20December%202020.pdf.

Daniels, T., Goodacre, L., Sutton, C., Pollard, K., Conway, S., & Peckham, D. (2011). Accurate assessment of adherence: self-report and clinician report vs electronic monitoring of nebulizers. Chest, 140(2), 425–432. https://doi.org/10.1378/chest.09-3074.

Eakin, M. N., Bilderback, A., Boyle, M. P., Mogayzel, P. J., & Riekert, K. A. (2011). Longitudinal association between medication adherence and lung health in people with cystic fibrosis. Journal of Cystic Fibrosis, 10(4), 258–264. https://doi.org/10.1016/j.jcf.2011.03.005.

Forbes, C. A., Deshpande, S., Sorio-Vilela, F., Kutikova, L., Duffy, S., Gouni-Berthold, I., & Hagström, E. (2018). A systematic literature review comparing methods for the measurement of patient persistence and adherence. Current Medical Research and Opinion, 34(9), 1613–1625. https://doi.org/10.1080/03007995.2018.1477747.

NHS England. (2014). Clinical commissioning policy: Inhaled therapy for adults and children with cystic fibrosis. https://www.england.nhs.uk/commissioning/wp-content/uploads/sites/12/2015/01/a01-policy-inhld-thrpy-cf.pdf.

Quittner, A. L., Zhang, J., Marynchenko, M., Chopra, P. A., Signorovitch, J., Yushkina, Y., & Riekert, K. A. (2014). Pulmonary medication adherence and health-care use in cystic fibrosis. Chest, 146(1), 142–151. https://doi.org/10.1378/chest.13-1926.

Ryan, C., Ross, S., Davey, P., Duncan, E. M., Francis, J. J., Fielding, S., Johnston, M., Ker, J., Lee, A. J., MacLeod, M. J., Maxwell, S., McKay, G. A., McLay, J. S., Webb, D. J., & Bond, C. (2014). Prevalence and causes of prescribing errors: The PRescribing outcomes for trainee doctors engaged in clinical training (PROTECT) study. PloS One, 9(1), e79802. https://doi.org/10.1371/journal.pone.0079802.

Sabaté, E. (2003). Adherence to long-term therapies: Evidence for action. World Health Organisation. https://apps.who.int/iris/bitstream/handle/10665/42682/9241545992.pdf.

Sawicki, G. S., Ren, C. L., Konstan, M. W., Millar, S. J., Pasta, D. J., Quittner, A. L., & Investigators and Coordinators of the Epidemiologic Study of Cystic Fibrosis (2013). Treatment complexity in cystic fibrosis: Trends over time and associations with site-specific outcomes. Journal of Cystic Fibrosis, 12(5), 461–467. https://doi.org/10.1016/j.jcf.2012.12.009.