Skip to main content

Patient-physician agreement among vulnerable ethnic minority patients in Denmark

Astrid Marie Høeg Næraa1, Dorthe Susanne Nielsen1, 2 & Morten Sodemann1, 2

14. okt. 2021
14 min.



Medical doctors in Denmark struggle to provide adequate treatment for vulnerable ethnic minorities [1]. Many diagnoses and essential problems are overlooked even after repeated healthcare contacts, potentially causing delayed or incorrect treatments and reducing trust in the healthcare system [1, 2].

Language barriers is one of the main reasons for this [1]. Another is the complex nature of patients’ problems including issues of health literacy, low socioeconomic status, multimorbidity and the prevalence of post-traumatic stress disorder (PTSD), etc. [1, 3]. Health is affected by social determinants such as short or no education, a low income and early loss of physical function [4, 5]. Furthermore, PTSD may impair the patient’s cognitive function and alter the way they perceive and describe their symptoms, which may in turn mislead doctors when facing clinical decisions [3]. An impaired cognitive function may result in problems maintaining a steady income or keeping track of official appointments, i.e. with a healthcare contact [2, 3].

Identifying these problems early on and co-producing healthcare with the patient may possibly help doctors find a solution with which the patient may comply [6, 7]. Thus, studies have indicated that doctors who create solutions together with their patients based on the patients’ needs and resources achieve more functional solutions. In contrast, doctors who provide solutions based on the needs and resources of the organisation tend to create dysfunctional solutions in which the patient is neglected [8, 9].

Currently, vulnerable ethnic minority patients are often hard to involve, and time pressure and language barriers steal time from patient engagement [7, 10]. A form of patient engagement coined co-production of carehas shown that this problem may be remedied if both the system and the professionals are adequately prepared [7].

Aiming to assess a possible target point, this study investigated one of the key elements in coproduction of care: the physician’s awareness of the patient’s priorities.

The purpose of this study was to assess the level of agreement between vulnerable ethnic minority patients and their doctors regarding the perception of the patients’ major health complaints.


This was a retrospective cross-sectional study based on the records of 150 patients referred to the Migrant Health Clinic in Odense, Denmark. Patients were referred to the clinic between 7 June 2016 and 31 December 2018, either by their general practitioner or by a hospital department in the Region of Southern Denmark.

Data were collected between 1 March and 1 October 2019.


The Migrant Health Clinic is a public hospital-based out-patient clinic receiving vulnerable immigrant and refugee patients with complex health problems who are referred by general practitioners and hospital departments. The aim is to help the patients on multiple levels, and the clinic employs both doctors, nurses and social workers. One of the essential clinical tools is a “Problem list”, an exhaustive list of patient-reported problems on which the patient is prompted to state all their problems regardless of their nature: medical, socioeconomic, psychological or other [11].

Only referral notes and notes from the two initial standard consultations were analysed in this study.


Only patients with at least the two standard consultations (including a “Problem list”) were eligible for this study. Patients were chosen randomly from a chronological list of Danish Civil Registration (CPR) numbers using a random number generator. The list only contained the CPR numbers and a unique patient identifier coined the study subject ID.


The primary outcome was the level of agreement between the referral and the “Problem list”, which is a list of problems coproduced with the patient when he or she first visits the Migrant Health Clinic [11].

The secondary outcome was the level of agreement between the referral and the full Migrant Health Clinic notes, including the “Problem list” and everything discovered during the two primary consultations at the clinic.

The tertiary outcome was to establish the number of patients with an overlap between one or more problems in the referral and the “Problem list” as well as the referral and the entire Migrant Health Clinic (MHC) note data material.

Data sources and methods

Data were extracted from referral notes and from Migrant Health Clinic records produced during the two initial consultations. The Migrant Clinic records were divided into the “Problem list” and any remaining text, where the patient answers specific questions from the staff, the latter simply being categorized as “Observed problems”. Each specific problem could appear only once per patient to avoid overrepresentation.

The “Problem list” is an established tool in the clinic and was deemed the best estimate of the patient’s experienced problems [12].


From the referral notes and clinic records, three main descriptive areas were defined under which patient information was allocated: medical problems, socioeconomic problems and socio-demographic background factors.

Every new problem encountered in the referral or clinic notes was transferred to a Microsoft Access (2016, version 16) table and assigned a unique number as a code identifier.

Medical problems included diagnoses and physical and mental health complaints as well as compliance-, pharmacology- and treatment-related problems. Socio-economic problems included all current socio-economic and emotional issues. Finally, demographic factors consisted of descriptions of the patient and his or her environment that were not presented directly as a problem but as factors describing the patient’s situation.

The codes were based on previous studies at the clinic and adjusted during the course of the study (before the data were analysed statistically) [13].

Only the corresponding author conducted the data collection. The supervisor made random quality assurance checks.


A minimum sample size of 88 patients was calculated using a 20% error margin and a 50% difference between the overall probability of agreement and the probability of agreement expected by chance alone, as suggested by Gwet [14].

To ensure a higher statistical power, a sample size of 150 patients was chosen before any statistical calculations were conducted as this was deemed possible within the time limit.

The chosen statistical software was RStudio Team (2018. RStudio: Integrated Development for R. RStudio, PBC, Boston, MA. V. 1.1.463)), STATA (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC) and Microsoft Excel (2010).

To describe the level of agreement between patient-perceived problems and the referring doctor’s perception, Cohen’s kappa coefficient (κ) and Chamberlain’s proportionate positive agreement (pppa) were calculated [15, 16]:

First, the number (%) of patient referral pairs with at least one matching problem in the referral and MHC notes was calculated (see Table 1). Next, it was determined how frequently each problem was reported in the MHC notes only, in the referral only and lastly in both the referral and clinic notes (see Table 1). The number was calculated for the “Problem list” and the total MHC notes, respectively.

No p-value was calculated since the null hypothesis is generally not applicable to kappa [17]. Instead, Chamberlain’s pppa was calculated since κ is problematic when the prevalence of an overlapping problem is low compared with the total number of times it is mentioned [15]. The pppa was read as a regular proportion [15].

Although a κ-value of 0.80 is often recommended as the minimum accepted value of agreement, this depends on the type of measurement [16]. In this study, a κ-value of 0.6 or above (corresponding to moderate, strong, or almost perfect overlap) or a pppa of 0.6 or higher was considered sufficient agreement because of the expected complexity in defining the patients’ problems.


Only patients who consented to participate in the research and had this stated explicitly in the patient files were eligible for the study.

Permission to handle personal data was granted by the Danish Data Protection Agency 2016-41-4693 and by The Region of Southern Denmark 19/7712.

Trial registration: not relevant.


Most patients were female, from Syria and had lived in Denmark for about 14 years (see Table 2). Only two patients did not require a translator. Often, both the patient and his or her partner were on social allowance (82% and 40%, respectively). Only 2% were currently working. Less than 50% had completed elementary school training and about 33% had worked in their home country and/or Denmark.

Only two codes were above the 0.6 cut-off in the primary outcome: musculoskeletal pain and Type 2 diabetes mellitus, and only musculoskeletal reached the 0.6 cut-off for pain in the secondary outcome (see Table 1).

In terms of the tertiary outcome, most patients had an overlap of one or more medical issues when comparing the referral to the “Problem list”, as well as the referral and the entire MHC data relating to the patient (93.33% and 97.33% respectively – see Table 1). When the patients had a wide range of problems, at least one patient-referral match for each patient may be expected. However, most problems were primarily mentioned in either the referral or the migrant notes and were rarely noted in both places for any single patient (see Figure 1).

In contrast, only 20% of the patients had an overlap of one or more socioeconomic issues when comparing the referral and the “Problem list”. This figure only rose to 43.33% when including all MHC data. A greater prevalence of socioeconomic issues such as trauma and financial difficulties was found in the clinic notes than in the referrals (see Table 3).


For the primary and secondary outcomes, almost no agreement was found between the referring doctors and the patients regarding the patients’ main problems. A considerable degree of overlap was found in the tertiary outcomes regarding the medical but notthe socioeconomic issues.

Choosing only one reviewer ensured consistency and easily comparable results and allowed us to avoid interrater variability. Determining the problems based on short texts only is largely subjective and might be done differently in a similar study. The use of interpreters and the fact that the patients’ problems were documented by a staff member are limitations that may potentially mean that important points were omitted or misunderstood. However, the “Problem list” was co-produced and reviewed in cooperation with the patient several times by both a physician and a nurse, making it a unique strength of this study.

The frequencies of some problems were very low, especially in the primary outcomes and socioeconomic problems in general. This was partly due to the narrow problem definitions chosen in order to avoid a falsely high degree of agreement. This makes meaningful interpretation of the results in those categories harder. Larger studies or broader categories may provide a greater validity.

Disagreement between referrals and patient-reported problems is not limited to this patient group, but establishing a universal solution covering all patients was beyond the scope of this study [18].

This study does not suggest that the patient’s viewpoint is the right one – merely that it is not in line with the referring doctor’s viewpoint. The clinic will most likely uncover more problems than described in the referral, which is why the comparison of the referral and the entire MHC notes were only the secondary outcome. However, the referral and the “Problem list” should both be a condensed version of essential problems regarding the patient. The difference between the two indicates that doctors and patients focus on different issues. Co-production of care has shown promising results in terms of aligning the two, but will most likely require rethinking parts of the healthcare system [7]. Perceiving that this is an easy task that will likely yield results rapidly may not be realistic. Nevertheless, the challenges described within the patient group are already an issue and will presumably remain so if no measures are implemented. Testing and validating even small interventions might prove to save time and avoid some futile treatments and workups. Implementing changes in general practice rather than hospitals may be the easiest way to start,as GPs will likely have highest number of patient encounters. Possible interventions aiming to achieve a higher agreement include awarding the general practitioners with a higher fee for more in-depth consultations with complex patients. This will compensate the general practitioners for their effort and, hopefully, help make the patient-perspective clearer to the referring doctors.

Easy access to qualified interpreters is a simple way of increasing the level of patient-physician understanding. Lastly, screening for known problems within the patient group, especially non-medical issues, may be an applicable alternative to the systematic questioning used at the Migrant Health Clinic, as well as converting the “Problem list” into a tool tailored for the needs of general practice, to name but a few suggestions [19]. Further studies are warranted to develop and validate interventions within this patient group.

If co-production of care is not currently possible in general practice, new criteria for referrals may be needed to better equip the receiving departments. Presently, the referral must be a short overview of the patient’s situation phrased by the doctor [20]. Perhaps the referral could include a part representing the patient’s perspective to ensure that their priorities are clearer to the general practitioner and presented more adequately to the receiving department.

Correspondence Astrid Marie Høeg Næraa. E- mail:
Accepted 7 September 2021
Conflicts of interest Potential conflicts of interest have been declared. Disclosure forms provided by the authors are available with the article at
Acknowledgements Stud.eocon. Serkan Korkmaz contributed with statistical calculations
References can be found with the article at
Cite this as Dan Med J 2021;68(11):A03210244



  1. Vikkelsø L, Al-zyadi H. [Equal access to health care - an analysis of GP's encounter with patients with ethnic minority background] [in Danish]. Lige adgang til sundhed – en analyse af praktiserende lægers møde med patienter med etnisk minoritetsbaggrund. Report. Copenhagen: Department for Equal Treatment, Danish Institute for Human Rights, 2014.

  2. Sodemann M. [The doctor's role in discrimination. What you don't know will hurt the patient] [in Danish]. Lægens rolle i forskelsbehandling. Det du ikke ved får patienten ondt af. Presentation. Odense, 2020:16-8.

  3. Sodemann M. [Language learning and post-traumatic stress disorder among refugees - an overlooked problem with health-related and legal consequences: a literature review] [in Danish]. Sprog indlæring og Posttraumatisk Stress Syndrom blandt flygtninge - et overset problem med helbredsmæssige og juridiske konsekvenser: En litteraturgennemgang. Presentation. Odense University Hospital, 2014.

  4. Diderichsen F, Andersen I, Manuel C. [Inequality in health - causes and efforts] [in Danish]. Ulighed i Sundhed - årsager og indsatser Copenhagen: Danish Health Authority, 2011.

  5. Bhattacharya J, Hyde T, Tu P. Socioeconomic disparities in health. Health economics. United Kingdom: Palgrave MacMillan, 2014:51-75.

  6. Batalden M, Batalden P, Margolis P et al. Coproduction of healthcare service. BMJ Qual Saf 2016;25:509-17.

  7. Radl-Karimi C, Nicolaisen A, Sodemann M et al. Under what circumstances can immigrant patients and healthcare professionals co-produce health? - an interpretive scoping review. Int J Qual Stud Health Well-being 2020;15:1838052.

  8. Simonsen LM. Interactional hybridity on the healthcare agenda. Dihm Digest 2018;8:4-5.

  9. Simonsen LM. [Framings in distributed cognitive systems: How practitioners meet the patient] [PhD thesis in Danish]. Framings i distribuerede kognitive systemer: hvordan praktikere møder patienten. University of Southern Denmark, 2017.

  10. Ahmed S. Experiences of communication barriers between physicians and immigrant patients: a systematic review and thematic synthesis. Patient Exp J 2017;4:122-40.

  11. Sodemann M. [A strategy for the conversation. What you don't know will hurt the patient] [in Danish]. En strategi for samtalen. Det du ikke ved får patienten ondt af. Odense, 2020:199-201.

  12. Sodemann M. Understanding unexplained and complex symptoms and diseases. In: Kumar BN, Diaz E, eds. Migrant health: a primary care perspective. CRC Press, 2019:171-84.

  13. Sodemann M. [Misunderstood symptoms, serious misconceptions and unintended incidents involving ethnic minority patients with language barriers - frequent challenges of patient safety involving language barriers] [in Danish]. Misforståede symptomer, alvorlige fejlopfattelser og utilsigtede hændelser involverende etniske minoritetspatienter med sprogbarrierer - hyppige udfordringer for patientsikkerheden ved sprogbarrierer Indvandrermedicinsk Klinik, Infektionsmedicinsk Afdeling, Odense Universitetshospital, 2017.

  14. Gwet KL. Sample size determination 2010. (21 Jan 2020).

  15. Cicchetti DV, Feinstein AR. High agreement but low kappa: II. Resolving the paradoxes. J Clin Epidemiol 1990;43:551-8.

  16. McHugh ML. Interrater reliability: the kappa statistic. Biochem Med (Zagreb) 2012;22:276-82.

  17. Kirkwood BR, Sterne JA. 36.3 Assessing reproducibility of measurements. Essential medical statistics. 2nd ed. Malden Massachusetts, USA: Blackwell Science, 2003:433-9.

  18. Vermeir P, Vandijck D, Degroote S et al. Communication in healthcare: a narrative review of the literature and practical recommendations. Int J Clin Pract 2015;69:1257-67.

  19. Noseworthy J. The future of care - preserving the patient-physician relationship. N Engl J Med 2019;381:2265-9.

  20. MedCom. ["The good referral"] [in Danish]. "Den gode henvisning". (14 Jan 2021).