CodeX effectively identifies high-risk patients with asthma or COPD in Dutch primary care,supporting guideline-driven treatment

Abstract

Background:

Prevention of lung attacks (LAs)/exacerbation is an important treatment goal in both asthma and chronic obstructive pulmonary disease (COPD). However, LAs are often not registered as such in medical records.

Objectives:

Development and evaluation of CodeX Asthma and COPD.

Design:

An electronic medical record-based algorithm to identify LAs in Dutch primary care patients with asthma or COPD was developed. The algorithms were evaluated in nine general practices in the Netherlands.

Results:

A total of 479 LAs (in 1164 patients) were identified with CodeX Asthma in the past year, of which only 16% were registered. CodeX COPD identified 321 LAs (in 242 patients) in the past 3 years, of which two were registered.

Conclusion:

CodeX algorithms are capable of identifying unrecorded LAs and high-risk/uncontrolled patients in an easy way. This offers primary care providers a simple solution to easily identify and closely manage high-risk patients with asthma or COPD by identifying LAs’ frequency and potential under- or overtreatment.

Introduction

Lung attacks (LAs), which are the renaming of exacerbations in the Netherlands, ¹ are common in patients with respiratory diseases. Among patients with chronic obstructive pulmonary disease (COPD) undergoing treatment in Dutch primary care, 34% have at least one lung attack/exacerbation per year, and 12% have two or more lung attacks/exacerbations per year. ² In patients with asthma, LAs may occur across all asthma severities and often despite optimal medical treatment. ³ Once a patient has experienced a lung attack, this increases the risk of future LAs.^4,5 In addition, LAs are a significant burden for both the patient and the healthcare system.^{6 –8} However, LAs are often not registered as such in the medical records. ⁹ Therefore, identification of LAs is important, and prevention of LAs is an important treatment goal in both asthma and COPD.^2,10,11

Quality improvement (QI) activities are crucial to improve the quality of care and optimize the management of patients with chronic diseases in primary care. ¹² These activities, such as risk prediction tools, can play a key role in preventing LAs by identifying at-risk patients requiring further follow-up. This will increase the proportion of patients receiving guideline-recommended treatment. ¹² The Netherlands based, but internationally operating General Practitioners Research Institute (GPRI) has illustrated its vision for healthcare improvement in primary care (Figure 1). These solutions could improve the patient’s pathway from detecting diseases to guideline-based management.

Figure 1.

GPRI healthcare improvement pathway.

The GPRI healthcare improvement pathway is divided into four phases: (1) the undiagnosed, (2) the diagnostic, (3) the diagnosed (but uncontrolled), and (4) the management phase. The diagnostic phase is currently being investigated in the Breathlessness diagnostics in a box (BiaB) study. ¹³ The main aim of this study is to demonstrate whether the use of BiaB will shorten the time to diagnosis as compared to usual care. Examples of QI tools in the management phase are the AsthmaOptimiser and COPDOptimiser (Figure 1). ¹⁴ These tools provide guideline-based management suggestions and provide structured guidance for healthcare professionals during consultations in primary care.

CodeX Asthma and CodeX COPD are two practical tools that can be used for the “Diagnosed (but uncontrolled)” phase. The CodeX tools are algorithms that quantify all LAs, that is, those registered as such in the electronic medical records (EMRs) using guideline-specified coding as well as unregistered LAs in patients with asthma or COPD in Dutch primary care, thereby identifying patients at high risk of LAs. The algorithm identifies these LAs using data from EMRs of the general practitioner’s (GP) practice. This data consists of presenting symptoms and prescription or change of (current) medication. This is of importance as LAs are often not registered in the EMR as such. ⁹ Therefore, no adequate follow-up is triggered. This may lead to unrecognized LAs and a lack of appropriate management or follow-up. In this article, the methodology and scientific results of the CodeX tools are described.

Methods

CodeX algorithms development

The first step of this QI project, including developing CodeX Asthma and CodeX COPD, was to define what constitutes a lung attack. This was described separately for asthma and COPD based on a literature search. Results of the literature search were discussed with international scientific experts (Scientific Advisory Board, co-authors of this manuscript). In addition, selected markers for a lung attack were checked for availability in the EMR.

During the pilot study, there was regular discussion between the study team and the scientific advisory board to evaluate and optimize the CodeX algorithm. For example, an indicator based only on prescribed medication has been removed since it was causing much noise in the report.

Patient-level report development

The indicators (Supplemental Table 1) for a lung attack in asthma or COPD were combined into an algorithm to identify unregistered LAs. The VIPlive system (Topicus, Deventer, The Netherlands) allowed to run the algorithm in each participating practice. This resulted in a patient-level report containing a list of all patients with registered (Dutch College of General Practitioners (NHG) code 3549 (number of severe LAs in past year) and NHG 3014 (new lung attack requiring OCS)) and unregistered LAs (based on the CodeX algorithm) in the past year and past 3 years. Additional information provided was prescribed (inhalation) medication before and after the lung attack, and time (days) until the first contact with the practice after a lung attack. Those reports are called CodeX Asthma and CodeX COPD.

CodeX Asthma includes all patients with an active episode of asthma (ICPC R96 or R96.2) aged 5 years and older, and CodeX COPD includes all patients with an active episode of COPD (ICPC R95) aged 40 years and older.

Pilot study

For each CodeX algorithm, a pilot study was performed to exploratively assess the value of the CodeX algorithms (CodeX Asthma: September 2023, CodeX COPD: February 2022). GP practices were invited by GPRI to participate in this study. Participating practices were either current users of VIPlive or were connected to VIPlive for this study. Practices were requested to download the patient-level report for their practice, and pseudonymized reports were shared with GPRI for study purposes. Afterward, the GP practices were asked for their opinion regarding the use of the algorithm.

Data analysis

For each of the CodeX algorithms, data from the participating practices were combined. Analyses were performed in R (version 4.3.3) for asthma and COPD separately and consisted of descriptive characteristics. No significance testing was performed. We report the prevalence of registered and unregistered LAs and the follow-up time since the last lung attack. For asthma, we stratify these results based on whether the patient was prescribed an inhaled corticosteroid (ICS).

The SQUIRE reporting guidelines were used for the preparation of this manuscript. ¹⁵

Results

CodeX algorithms development

The defined indicators for a lung attack are a combination of registered LAs, according to the Dutch guideline coding (NHG 3014), and indicators considered to reflect a lung attack that are not registered as such. This resulted in eight indicators for asthma and six indicators for COPD (Supplemental Table 1).

Pilot study

The asthma report was evaluated in five GP practices with a total number of 13,192 patients. From this group, 1164 (9%) patients, aged 5 or over had an active asthma episode. To evaluate the COPD report, four GP practices were included with a total number of 242 patients with an active episode of COPD. Initially, five GP practices joined the study but one dropped out due to technical difficulties with VIPlive.

Results CodeX Asthma

Within the 1164 patients with asthma, 288 (25%) patients experienced at least one lung attack in the past year, resulting in a total of 479 LAs registered by the algorithm. Of all 479 LAs, only 60 (16%) were registered as such in the EMR with Dutch NHG codes. Supplemental Table 2 shows the baseline characteristics of the patients with asthma, stratified on patients who had not experienced a lung attack compared to patients who had at least one lung attack in the past year identified by the algorithm.

A total of 479 (41%) patients with asthma were treated with ICS, of those 172 (36%) experienced at least one lung attack in the past year, of which 52 (30%) had two or more LAs (Figure 2(a)). A total of 685 (59%) patients were not treated with ICS, of those 116 (17%) had at least one lung attack in the past year, of which 20 (17%) had two or more LAs. Of the patients with a lung attack respectively 59 (34%) and 34 (29%) patients had a follow-up appointment within 4 weeks after their last lung attack.

Figure 2.

Results of CodeX Asthma and COPD in respectively 5 and 4 Dutch GP practices. Number of patients with lung attacks in the past year according to the CodeX algorithm and follow-up after lung attack is shown. Results for CodeX Asthma are stratified based on ICS use. Data is shown as n (%). (a) CodeX Asthma and (b) CodeX COPD.

A total of 116 (40%) patients with a lung attack were not treated with ICS nor started ICS after the attack. A total of 511 (44%) patients registered as having active asthma were not treated with inhalation therapy nor experienced an attack in the past year.

Results CodeX COPD

Within the 242 patients with COPD, 99 (41%) patients were identified to have had a lung attack in the previous 3 years, resulting in a total of 321 LAs registered by the algorithm. Only two of these 321 (0.6%) LAs were registered as such in the EMR with Dutch NHG codes.

When stratifying these 99 patients over the past year, (a) 65 (66%) of them had no LAs in this past year (Figure 2(b)), (b) 19 (19%) patients had one lung attack, and (c) 15 (15%) patients had two or more LAs. Patient characteristics are provided in Supplemental Table 3. The median follow-up time after a lung attack, in patients with a single lung attack, was 3 weeks, with 8 patients not having a follow-up appointment within 4 weeks. The median follow-up time after a lung attack, in patients with two or more LAs, was within 1 week, with 3 patients not having a follow-up appointment within 3 weeks.

User feedback for both the asthma and COPD algorithms indicated an easy-to-use tool that is helpful in identifying unregistered LAs.

Discussion

CodeX algorithms for asthma and COPD were developed and implemented in Dutch EMR systems to identify unregistered LAs and high-risk or uncontrolled patients with asthma or COPD needing additional care.

The majority of LAs identified by the algorithm were not registered as such in the EMR. This could lead to lower awareness of the treating healthcare professional about the severity of a patient’s asthma or COPD. For patients with COPD, the strongest predictor of a next lung attack is the number of LAs they have had in the prior year. ¹⁶ For patients with asthma, having ongoing LAs despite correct inhaler technique and good medication adherence is an indication for referral to secondary care. ¹⁷

According to international guidelines adults, adolescents, and children with asthma should receive ICS-containing treatment to reduce their risk of serious LAs and to control symptoms. ¹⁷ We observed in this study that only 41% of the patients with an active diagnosis of asthma according to the EMR received ICS-containing treatment. Of the patients not receiving ICS-containing treatment, 17% had a lung attack according to the algorithm in the past year. The use of tools such as CodeX can assist a healthcare professional in identifying patients who need a review of their disease management including medication. On the other hand, 83% of the patients not treated with ICS did not have a lung attack in the past year, suggesting a review of their asthma diagnosis might be indicated.

The CodeX algorithms are part of GPRI’s healthcare improvement pathway. Patients identified with CodeX can further be assessed using tools such as the AsthmaOptimiser or COPDOptimiser. Those tools provide guideline-based management suggestions and provide structured guidance for healthcare professionals during consultations in primary care. Assessing the number of past LAs is part of this assessment. The AsthmaOptimiser is registered as a Class IIa medical device in Europe and can be downloaded by Dutch healthcare providers at http://www.asthmaoptimiser.com, the COPDOptimiser is currently being developed.

A potential limitation of CodeX is the fact that the indicators are based on alternative diagnostic codes such as pneumonia and acute bronchitis. It could be the situation in which a patient with asthma or COPD has acute bronchitis and that this is recorded correctly in the EMR, instead of an asthma/COPD unregistered lung attack. However, since these symptoms are an important risk factor for LAs in asthma and COPD it is of great value for a healthcare professional to have more insight into this.^18,19 In addition, the awareness of these patients at risk by the healthcare provider is one of the aims, more than the lowest possible false positive cases.

The CodeX Asthma report used in the pilot study did not distinguish moderate (requiring inhaled medication change) from severe (requiring a course of antibiotics+/− oral corticosteroids) LAs. In the next version of the report this will be incorporated, classifying LAs in combination with a prescription of systemic corticosteroids or antibiotics as a severe lung attack and LAs without a medication prescription or a prescription or change in inhaler therapy as moderate LAs. This additional level of information might be useful in the management of the disease. ¹⁷

The intention of this project was to develop tools based on literature and expired opinions to be implemented and improved in Dutch primary care, rather than conducting a diagnostic accuracy study. The current report describes the development of the tool and the capabilities of identifying unregistered LAs. However, no clinical follow-up has been performed to assess the effects on patients’ outcomes after identifying unregistered LAs. The algorithms are now implemented in multiple practices in the Netherlands and healthcare groups are requesting access suggesting interest in the tools. In August 2024, over 200 GP practices had access to the tools.

We have developed the algorithms to detect LAs for the Dutch general practice EMR systems, using the International Classification of Primary Care (ICPC) coding for diagnosis and the ATC codes for medication. These codes can be mapped to other coding systems to be used in another setting for a wider implementation. ²⁰

Conclusion

With CodeX we developed and implemented two useful tools for primary care. CodeX Asthma and COPD can identify unrecorded lung attacks and high-risk/uncontrolled patients in an easy way. This will help primary care providers manage high-risk patients by tracking lung attack frequency and treatment adequacy, thereby supporting guideline-directed care.

Supplemental Material

sj-docx-1-tar-10.1177_17534666251329192 – Supplemental material for CodeX effectively identifies high-risk patients with asthma or COPD in Dutch primary care, supporting guideline-driven treatment

Supplemental material, sj-docx-1-tar-10.1177_17534666251329192 for CodeX effectively identifies high-risk patients with asthma or COPD in Dutch primary care, supporting guideline-driven treatment by Iris van Geer-Postmus, Marika T. Leving, Yoran H. Gerritsma, Esmé Baan, Lars Dijk, Evelien Harms, David Price, Gerian H. Prins, Jennifer K. Quint, Dermot Ryan, Philippe Salomé, Björn Ställberg, Nilouq Stoker and Janwillem H. Kocks in Therapeutic Advances in Respiratory Disease