Article Text
PDF
Abstract
Objectives The existing evidence for the impacts of continuity of care (COC) in patients with chronic obstructive pulmonary disease (COPD) is low to moderate. This study aimed to investigate the associations between relational COC within primary care and COPD-related hospitalisations using a robust methodology.
Design Population-based cohort study.
Setting National Health Insurance Service database, South Korea.
Participants 92 977 adults (≥40 years) with COPD newly diagnosed between 2015 and 2016 were included. The propensity score (PS) matching approach was used. PSs were calculated from a multivariable logistic regression that included eight baseline characteristics.
Exposure COC within primary care.
Main outcome measures The primary outcome was the incidence of COPD-related hospitalisations. Cox proportional hazard models were used to estimate HRs and 95% CIs.
Results Out of 92 977 patients, 66 677 of whom were cared for continuously by primary doctors (the continuity group), while 26 300 were not (the non-continuity group). During a 4-year follow-up period, 2094 patients (2.25%) were hospitalised; 874 (1.31%) from the continuity group and 1220 (4.64%) from the non-continuity group. After adjusting for confounding covariates, patients in the non-continuity group exhibited a significantly higher risk of hospital admission (adjusted HR (aHR) 2.43 (95% CI 2.22 to 2.66)). This risk was marginally reduced to 2.21 (95% CI 1.99 to 2.46) after PS matching. The risk of emergency department (ED) visits, systemic corticosteroid use and costs were higher for patients in the non-continuity group (aHR 2.32 (95% CI 2.04 to 2.63), adjusted OR 1.25 (95% CI 1.19 to 1.31) and expβ=1.89 (95% CI 1.82 to 1.97), respectively). These findings remained consistent across the PS-matched cohort, as well as in the sensitivity and subgroup analyses.
Conclusions In patients with COPD aged over 40, increased continuity of primary care was found to be associated with less hospitalisation, fewer ED visits and lower healthcare expenditure.
- COPD epidemiology
Data availability statement
Data may be obtained from a third party and are not publicly available. The raw data that support the findings of this study are available only for authorised researchers in South Korea and for a limited period due to the information protection law for patient privacy. Study protocol is available from the corresponding author on request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Clinical and economic outcomes improve with high levels of relational continuity between doctors and patients.
However, there is limited empirical evidence demonstrating the impact of continuous primary care on chronic obstructive pulmonary disease-related hospitalisations.
WHAT THIS STUDY ADDS
This population-based study, with internal validity, demonstrated that reduced continuity of care in primary settings correlated with a heightened risk of hospital admission.
The findings, investigated in a healthcare system with markedly different environments, would enhance the external validity of the evidence.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
This study highlights the need for further investigation into how factors influencing continuity levels vary according to different healthcare system environments.
Healthcare providers may need to prioritise strategies that enhance ongoing patient–provider relationships.
The evidence supports the development of policies that promote consistent primary care relationships.
Introduction
It is widely acknowledged that primary care quality efficiently improves overall health outcomes,1 2 and continuity of care (COC) is considered a core element of primary care.3–5 According to a systematic review by Huntley et al, COC is one of four identified systemic features that affect unscheduled secondary care utilisation, and the other three were access, practice features and quality of care.6 COC has been characterised by informational, management and relational continuity.2 7–9 Informational continuity involves the exchange of medical histories among current and previous healthcare providers. Management continuity entails coordinating and harmonising care according to shared treatment plans across providers.1 2 7 9 Relational continuity (also called interpersonal continuity) concerns long-term therapeutic relationships between patients and their healthcare providers.2 While informational and management continuity is an issue among service providers, relational continuity centres on the relationship between the service provider and the patient. Relational continuity places focus on the individual rather than on the illness, this concept reflects the value of primary care.1 A long-term patient–provider relationship can improve communication and establish trust,1 2 10 which results in a greater willingness to share crucial information with providers and increases the likelihood of patients adhering to treatment and preventive advice.10 These relationships also help healthcare providers understand their patients better, improve the effectiveness of chronic condition management and the development of long-term disease-monitoring strategies.1 10 Undoubtedly, high-quality primary care can facilitate planned end-of-life care in the community rather than in hospital.11 As a result, studies show that high relational COC might lower the risk of premature mortality and prevent the exacerbation of chronic conditions.12–16 Furthermore, patients with higher relational continuity tend to use secondary care less16–20 and lower medical costs.16 21 22 Conversely, low relational continuity negatively affects patient experience.23 24
Although evidence is accumulating for a wide range of conditions, evidence linking the benefits of high relational continuity in patients with specific conditions is insufficient. A recent systematic review expressed concern that the current level of evidence is low to moderate for asthma and chronic obstructive pulmonary disease (COPD).12 After this systematic literature review, two new studies on COC in patients with COPD have been published.25 26 However, the Canadian study still uses a design that measures exposure and outcome simultaneously, leading to confusion about the temporal relationship between cause and effect.25 Additionally, concerns have been expressed that it will become difficult to maintain patient–doctor relationships as complexity increases in terms of illnesses and medical organisations.4 27
Approximately, 97% of the Korean population is covered by the National Health Insurance (NHI) while the remaining 3% are covered by a medical aid programme (MedAid) that provides more comprehensive coverage to low-income households.28 The South Korean healthcare system is composed of primary, secondary and tertiary institutions. A mandatory referral document is required only for accessing tertiary hospitals. Tertiary hospitals are mostly university-affiliated institutions located in a few metropolitan cities. They serve as major centres for medical education and training, offering a wide range of specialised medical treatments. In the absence of legal restrictions, except for those involving tertiary hospitals, patients often prioritise proximity to medical facilities. Patients may receive primary care at clinics or hospital outpatient departments if these are within their catchment area. As a result, the country has been evaluated to have a weak gate-keeping primary care function.29 In this regard, the relationship between COC and outcomes in the Korean healthcare system has implications for other countries concerned about the weakening of primary care systems.
Set against the presented background, this study aimed to investigate the impacts of relational COC between patients with COPD and primary care doctors on clinical and economic outcomes. COPD is an ambulatory care sensitive condition (ACSC), and its effective management and treatment within ambulatory settings obviate hospital admission.30 31 We hypothesised that high relational care continuity with a primary care doctor results in better clinical results and lower costs.
Methods
Study design and data source
This population-based cohort study was performed according to the Strengthening the Reporting of Observational Studies in Epidemiology guideline.32 Levels of patient relational COC with their doctors were explored, comparative cohorts based on COC levels were established and the associations between COCs and hospital admission rates and other outcome measures were investigated. Anonymised national insurance claims data between 2014 and 2021 provided by the Korean National Health Insurance Service (KNHIS) were analysed. The KNHIS database contains details of all claims made by Korean residents. These details include deidentified patient sociodemographic information, diagnoses, all medical services provided and medications dispensed and death records.33
Study time frame
Figure 1 illustrates the study period from 2014 to 2021. Index dates were defined as the dates of the first diagnosis of COPD during 2015–2016. The 12-month period preceding the index date was defined as the preindex year. The exposure period was defined as the 2-year period following the index date, and the outcome period was defined as the period from the end of the exposure period to 31 December 2021. Patients were followed from the end of the exposure period until outcome determinations, death or the end of data collection, whichever came first.
Defined study periods.
Study population
Patients with COPD newly diagnosed in 2015–2016 at primary clinics, aged ≥40 years on index date, and who made at least four ambulatory visits during the exposure period were enrolled in the study.19 34 COPD was identified using the International Classification of Disease, 10th revision (ICD-10) as diagnosis codes J42–44 but excluding McLeod syndrome (a genetic disease coded J43.0). Patients with a claims record for COPD in the preindex year were excluded, and those who visited tertiary facilities, died or experienced hospitalisations or emergency department (ED) visits during the exposure period were also excluded. After exclusions, the final analytic cohort contained 92 977 unique patients.
Measuring continuity of primary care
Continuity of primary care is conceptualised as the extent to which a patient’s visits are concentrated among primary care doctors. Primary care is defined differently across countries.35 We defined primary care as community-based care and included visits to small-sized and medium-sized facilities, excluding tertiary hospitals, in the COC measurement. To measure COC, we used the Bice & Boxerman Continuity of Care Index (COCI) and Usual Provider Continuity (UPC).21 36 37 COCI is a dispersion index, and a useful metric for patients who may potentially visit different providers and is used chiefly for claims data analysis.38 COCI is recommended in South Korea because patients are almost free to contact doctors of choice due to the weak gatekeeper role of primary care.29 In this study, we also calculated UPC because it is one of the most popular continuity indices used in studies.38 UPC is a density index that quantifies patients’ visit patterns by focusing on specific providers.38
Bice and Boxerman COCI and UPC were calculated using the following formulae.37–39
Where N is the total number of visits made by a patient to a doctor, ni is the number of patient visits with provider i and M is the number of potentially available providers. In the UPC equation, provider i is the provider patients usually visit.
Both indices have a value of unity (1) if a patient always visits one specific health provider and a value of 0 if the patient visits different providers at each visit.
Outcome measures
The primary outcome measure was the incidence of COPD-related hospitalisation during the outcome period because hospitalisation rates for ACSCs can indicate the quality of primary care. Secondary outcomes were the incidence of COPD-related ED visits and COPD-related costs. Patients using corticosteroids in oral or injection form, defined as systemic corticosteroid use, were also investigated to evaluate COPD exacerbation. COPD-related medical costs were defined as the average annual medical costs for each year during the outcome period. Data related to COPD were identified by the ICD-10 diagnosis codes mentioned above. All medical costs were standardised costs to 2020 KRW to remove any effects of inflation.
Covariates
Covariates included individual characteristics such as age and sex. NHI contributions were classified as high, moderate or low and used as proxies of patients’ economic circumstances. Two types of health insurance programmes, that is, NHI and MedAid were included. Locations of residences at index dates were classified as large urban (metropolitan cities), small urban (other cities) or rural areas. Large and small urban, and rural areas were classified based on population densities. Elixhauser Comorbidity Indices (ECIs) were computed as proxies of patient health statuses40 based on diagnoses obtained from outpatient and inpatient records during the preindex year. In addition, we considered whether patients had a diagnosis of asthma, allergic rhinitis, atopic dermatitis, hypertension, diabetes, dyslipidaemia, cancer, osteoarthritis or rheumatoid arthritis during the preindex year or were being prescribed inhalers for COPD treatment at index dates or systemic corticosteroids during the exposure period. Patients who made more than 14 outpatient clinic visits during the exposure period were defined as frequent doctor visitors (90th percentile or higher). Based on index dates, disability status was defined as none, moderate or severe, and smoking status as never-smoker, ex-smoker, smoker or no record. For reference purposes, we also collected average annual COPD-related medical costs during exposure periods.
Configuration of the continuity cohort
Interim analysis was conducted to determine continuity score distributions of primary care doctor visits during the exposure period. Interim analyses showed that patients with a COCI of 1 accounted for 71.7% of doctor visits and those with a UPC of 1 accounted for 71.7%. We allocated patients with a continuity score of 1 during the exposure period to the continuity group and the others to the non-continuity group. Propensity score (PS) matching was used to improve comparability between the continuity and non-continuity groups and to reduce the effect of known confounders on study outcomes. PS matching was performed using a logistic regression model containing age, sex, health insurance programme, insurance contributions, residence urbanisation level, disability, ECI score and asthma coexistence as covariates. The continuity group was matched 1:1 to the non-continuity group without replication using the nearest matching method with 0.2 times the logit of PS as a calliper.
Statistical analysis
Baseline characteristics and healthcare utilisation were summarised using means, SD, medians and IQR for continuous variables, and frequencies and proportions for categorical variables. Intergroup differences in baseline characteristics between the continuity and non-continuity groups were assessed using the absolute standardised difference method and when they exceeded 0.1, the intergroup difference was considered as a meaningful difference between the two groups existed.41 COCIs, UPCs and annual health service utilisations during the exposure period in the two groups were compared by the independent t-test or the χ2 test, based on type of data. The effects of COC on hospitalisation and ED visits due to COPD were assessed using a Fine-Gray subdistribution hazard model that considered death as a competing risk.42 HRs and 95% CIs were provided. The use of systemic corticosteroids was assessed using a logistic regression model, and the results are presented as ORs. COPD-associated costs were estimated using a gamma regression model, and results are reported as exponential coefficients. In the analysis of the total cohort, sex, age, insurance programme, insurance contributions, urbanisation level of residence, disability, ECIs, asthma coexistence, COPD inhaler use, systemic corticosteroid use during the exposure period, smoking and frequent doctor visits were adjusted as covariates, and in the PS matched cohort, COPD inhaler use, corticosteroid use during the exposure period, smoking and frequent doctor visits were adjusted as covariates. Individuals whose information about insurance contribution, living area and smoking status was missing were not excluded, and missing values were considered as an independent category in the statistical models. A cumulative incidence graph was used to compare the continuity and non-continuity groups. For sensitivity analysis, we calculated COCIs every 2 years following the index date to construct longitudinal data. These data were then analysed using the generalised estimating equations method, adjusting for the same baseline characteristics and covariates as in the adjusted model for the total cohort. Subgroup analyses were performed on basic characteristics. The analysis was conducted by using SAS V.9.4 (SAS Institute), and statistical significance was accepted for p values <0.05.
Patient and public involvement
Patients and/or the public were not involved in the design, or conduct, or reporting or dissemination plans of this research.
Results
Baseline characteristics of the study population
The selection process is shown in figure 2. A total of 92 977 patients were eligible for analysis.
Identification of study participants and formation of study cohort. COCI, Continuity of Care Index; COPD, chronic obstructive pulmonary disease; ECI, Elixhauser Comorbidity Indices; PS, propensity score.
Table 1 shows the baseline characteristics of the study population. After classifying patients according to the COCI criteria, 66 677 were allocated to the continuity group and 26 300 to the non-continuity group. In the continuity group, there were more female patients than in the non-continuity group. Patients in the non-continuity group were older, had more beneficiaries, more with disabilities, more who were former or current smokers, more who had five or more chronic conditions and a higher copresence of asthma than those in the continuity group. In the non-continuity group, patients were less likely to live in large urban areas. 25 866 patients in the non-continuity group were matched to an equal number of patients in the continuity group by PS matching. The characteristics of matched cohorts were well balanced, showing no meaningful differences in major factors (table 1). In the exposure period, the non-continuity group made more frequent doctor visits and had higher medical expenses compared with the continuity group (table 1).
Baseline characteristics of study population and health service utilisation during the exposure period
Changes in COC
The mean COCI of study patients was 0.87 during the exposure period, which slightly decreased to 0.84 during the 4-year study outcome period. By group, while the mean COCI decreased from 1.0 to 0.96 in the continuity group, it remained around 0.54 in the non-continuity group throughout the study period (online supplemental figure 1). COCI changes were similar in the PS-matched cohorts (online supplemental figure 1). The mean UPCs were slightly larger than those of COCIs, and the pattern of changes was similar to that of the COCI (online supplemental figure 1). The average study periods for the continuity and non-continuity groups were 5.89 (SD 0.88) and 5.74 (1.05) years, respectively, and in the PS-matched cohort were 5.82 (0.96) and 5.74 (1.05), respectively.
Supplemental material
Risks of hospital admission
874 patients (1.31%) were hospitalised in the continuity group, and 1220 patients (4.64%) in the non-continuity for COPD-related reasons (table 2). After adjusting for key covariates, patients in the non-continuity group were found to be at significantly higher risk of hospitalisation (adjusted HR 2.43 (95% CI 2.22 to 2.66)). Figure 3 presents the cumulative incidence of hospitalisation during the outcome period from the primary analysis. The sensitivity analysis indicated a reduced but consistent result (adjusted HR 1.25 (95% CI 1.14 to 1.36)). In the PS-matched cohort, adjusted HR for hospital admission was 2.21 (95% CI 1.99 to 2.46). The demographic features that increased the risk of hospitalisation included being men, aged 65 or older, residing in small urban or rural areas, using a COPD inhaler at the index date, using systemic corticosteroids during the exposure period, having coexistence asthma, being a past or current smoker and being a frequent doctor visitor (online supplemental table 1).
Cumulative incidence of hospital admission and ED visit by level of continuity. (A) Hospital admissions in the total cohort, (B) Hospital admissions in the PS-matched cohort. (C) ED visits in the total cohort, (D) ED visits in the PS-matched cohorts. ED, emergency department; PS, propensity score.
Descriptive summary of study outcomes and association of continuity of primary care with study outcomes
Risks of ED visits
Patients in the non-continuity group were more likely to visit an ED than those in the continuity group (adjusted HR 2.32 (95% CI 2.04 to 2.63)) (table 2). However, the magnitude of the adjusted HR insignificantly decreased to 1.13 (95% CI 0.99 to 1.28) in the sensitivity analysis (p=0.061). In the PS-matched cohort, the adjusted HR for an ED visit was 2.06 (95% CI 1.78 to 2.38). The demographic features that increased the risk of an ED visit were similar to those of hospital admission (online supplemental table 1).
Risk of systemic corticosteroid use
6461 (9.69%) patients in the continuity group and 3807 (14.48%) in the non-continuity group were newly prescribed systemic corticosteroids (table 2). After adjusting for covariates, the non-continuity group had a 1.25-fold higher risk of being prescribed systemic corticosteroids (adjusted OR 1.25 (95% CI 1.19 to 1.31)). The adjusted OR of the PS-matched cohort was similar (1.29 (95% CI 1.22 to 1.37)).
COPD-related medical costs
Mean annual COPD-related medical cost was significantly higher in the non-continuity group (table 2). After adjusting for major covariates, gamma regression modelling analysis demonstrated patients in the non-continuity group spent on them about 1.89 (95% CI 1.82 to 1.97) times more than patients in the continuity group. The intergroup difference was similar for the PS-matched cohort (1.79 (95% CI 1.70 to 1.87) times more).
Subgroup analyses on hospitalisation and ED visits
Primary analysis trends were similar for all subgroups of the total cohort or PS-matched cohort (online supplemental figure 2).
Discussion
This study analysed nationwide claims data to investigate the impact of continuity of primary care on the clinical and economic outcomes of patients with COPD in the Korean healthcare environment. Study patients with low-level COCs with doctors were twice as likely to be hospitalised or visit an ED and incurred 1.8 times higher healthcare costs. In addition, the risk of receiving systemic corticosteroids due to COPD exacerbation was 1.25 times greater in the non-continuity group than in the continuity group, and similar results were obtained after PS matching. Sensitivity and subgroup analyses produced consistent results. The study shows that high relational continuity of primary care is associated with better health outcomes and lower healthcare costs, which concurs with previous studies.12 Recently, a study conducted in Ontario, Canada, reported that patients with COPD aged over 35 who received non-continuous primary care had a 2.81-fold higher risk of hospitalisation and a 2.12-fold higher risk of emergency room visits.25 Additionally, a Norwegian study found that patients in the lowest 30% for UPC had a 3.3-fold increased risk of death compared with those with a UPC of 1.26
This study contributes as the robust methodology employed enhances the internal validity of the findings from previous studies. We analysed nationwide claims data for the entire Korean population. PS matching increased comparability between cohorts. In order to avoid overestimating outcome risks, the period measuring continuity and outcomes was separated and sensitivity analysis was conducted.12 34 The analysis conducted on the PS-matched groups and the sensitivity analysis consistently yielded the same results, which underscored the reliability and validity of the study. Nonetheless, several limitations should be considered when interpreting the results of this study. First, the COC measurements used for the analysis were related to visit patterns and concentration rather than the interpersonal nature of the continuity relationship.7 Second, we used claims data for the analysis, which inherently introduces its shortcomings. For example, the claims data did not include information about individual doctors, which can be more problematic in a hospital outpatient setting than in a clinic. To address this, we operationally defined visits to the same institution and the same department in the hospital outpatient setting as visits to the same provider. In the Korean healthcare reservation system, patients are typically scheduled to see the same doctor if they visit the same medical department within the institution. Another important shortcoming is that we were unable to incorporate objective parameters reflecting patient health statuses (eg, laboratory results) into the analysis. However, the use of systemic corticosteroids was included as a secondary outcome measure to assess disease exacerbation. Third, the results of this study should be interpreted with the understanding that they pertain to patients with COPD in the early stages of the disease. Future research should also evaluate the impact of COC on end-of-life management in COPD. Lastly, care should be taken when generalising our results because the primary care facilities of this study were chosen to reflect the Korean situation. While the inclusion criteria were limited to primary clinics, visits to small-sized and medium-sized facilities, excluding tertiary hospitals, were considered in the measurement of COCI. This might limit the comparability of this study with previous research. However, our results were obtained from a different healthcare setting and are concordant with those of previous studies. In this regard, our analysis may contribute to improving the external validity of this theme.
Interestingly, despite the nearly free choice of healthcare providers in the Korean healthcare system, the proportion of study patients with a COCI of 1 was high at 72%. A COCI of 1 indicates that patients exclusively visited one doctor throughout the observation period, rendering UPC-based analyses meaningless, as a COCI of 1 is equivalent to a UPC of 1. The unexpectedly high COCI value among Korean patients with COPD contrasts with approximately 46% of Korean patients with newly developed dyslipidaemia16 and 62% of Austrian patients with diabetes who had a COCI of 1.43 However, direct comparisons of COC measurements across diseases or countries are challenging. For instance, in Norway, the proportion of patients with a UPC≥0.75 were lower in asthma (42%–48%) or COPD (49%–52%) than in diabetes (59%) or heart failure (62%–72%),44 which is the opposite of what was presented earlier. A recent study reported that COCI among patients with COPD in Norway was observed to be lower than that in Germany, even though Norway has a mandated gatekeeping system.45 This suggests that various factors, beyond health policy, could influence COC. Although there is growing awareness of a positive association between COC and clinical results, understanding of the factors that influence COC is limited.
In this regard, further research is needed to determine the reasons for the high COC exhibited by Korean patients with COPD. One possible explanation may lie in the nature of COPD, which is characterised by symptoms such as persistent, progressive airflow limitation. As compared with other chronic conditions like dyslipidaemia and early hypertension, COPD can cause immediate discomfort and increase patient desire for medication. Moreover, there is a high probability that a patient will return to the same doctor when symptoms are adequately managed using medication. In this regard, one US study reported that COCIs for COPD were slightly higher than those for congestive heart failure or diabetes.21 However, this might not be the sole reason for the high COC, given that in Taiwan, only 46% of patients with COPD exhibited a COCI of 1, even though the inclusion criteria in the Taiwanese study were similar to those in the present study.22 The high proportion of patients with a COCI of 1 in this study may reduce comparability with other studies. Due to this high proportion, classifying all cases with a COCI less than 1 as non-continuity suggests that our non-continuity group may actually have more continuity than those in other studies, potentially undervaluing the impact of primary care continuity.
This study’s findings offer crucial insights for policy-makers and primary care providers. Despite evolving patient–doctor dynamics, it is vital for primary care providers to sustain ongoing relationships with patients. In healthcare systems without traditional family doctor models, efforts to foster consistent connections between patients and their doctors are essential. Exploring options such as proactive information sharing can address challenges linked to multiple doctor involvement. Additionally, coordinated care among healthcare professionals may ease the complexities of fragmented encounters.
Data availability statement
Data may be obtained from a third party and are not publicly available. The raw data that support the findings of this study are available only for authorised researchers in South Korea and for a limited period due to the information protection law for patient privacy. Study protocol is available from the corresponding author on request.
Ethics statements
Patient consent for publication
Acknowledgments
This study analysed claims data from the Korean National Health Insurance Service (research management number NHIS-2023-1-211).
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
I-HL and EJJ contributed equally.
Contributors I-HL, EJJ, NKJ and EC wrote the statistical analysis plan, analysed data and SK carried out data cleaning and statistical analyses. I-HL and EJJ wrote the first draft of the manuscript. AJJ, EC and SK prepared tables and figures. NKJ and AJJ critically reviewed and edited the first draft. EJJ supervised all statistical analysis processes. I-HL was a fund holder and supervised all process of this research. All authors read and approved the final manuscript and met the ICMJE criteria for authorship. I-HL is the guarantor.
Funding This work was supported by a National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) grant number (2022R1F1A1073485).
Disclaimer The funding organisations had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Author note Transparency declaration: The lead author (I-HL) affirms that this manuscript is an honest, accurate and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.