Original Research | Published: 3 June 2024
open-url

Impact of patient–family physician language concordance on healthcare utilisation and mortality: a retrospective cohort study of home care recipients in Ontario, Canada

https://doi.org/10.1136/bmjph-2023-000762

Request reuse permissionopen-url
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) licenseopen-url

Abstract

Introduction As the world’s linguistic diversity continues to increase at an unprecedented rate, a growing proportion of patients will be at risk of experiencing language barriers in primary care settings. We sought to examine whether patient–family physician language concordance in a primary care setting is associated with lower rates of hospital-based healthcare utilisation and mortality.

Methods We conducted a population-based retrospective cohort study of 497 227 home care recipients living in Ontario, Canada. Patient language was obtained from home care assessments while physician language was obtained from the College of Physicians and Surgeons of Ontario. We defined primary care as language concordant when patients and their rostered family physicians shared a mutually intelligible language, and we defined all other primary care as language discordant. The primary outcomes were Emergency Department (ED) visits, hospitalisations and death within 1 year of index home care assessment.

Results Compared with non-English, non-French speakers who received language-discordant primary care, those who received language-concordant primary care experienced fewer ED visits (53.1% vs 57.5%; p<0.01), fewer hospitalisations (35.0% vs 37.6%; p<0.01) and less mortality (14.4% vs 16.6%; p<0.01) during the study period. In multivariable regression analyses, non-English, non-French speakers had lower risks of ED visits (adjusted hazard ratio [aHR] 0.91, 95% CI 0.88 to 0.94), hospitalisations (aHR 0.94, 95% CI 0.90 to 0.98) and death (aHR 0.87, 95% CI 0.82 to 0.93) when they received language-concordant primary care. For francophones, the risk of experiencing an ED visit, a hospitalisation or death was not impacted by the language of their family physician.

Conclusions Patient–family physician language concordance is associated with a lower risk of adverse outcomes in non-English and non-French speakers. Optimising the delivery of language-concordant care could potentially result in significant decreases in the use of acute healthcare services and mortality at the population level.

What is already known on this topic

  • Patients who face language barriers often face barriers to accessing healthcare services and also receive healthcare services of lower quality and safety.

  • However, it is not known whether these disparities can be attributed to cultural and/or language barriers, or whether outcomes can be improved by providing patients with language-concordant care.

What this study adds

  • Among a cohort of home care recipients living in Ontario, Canada, those whose primary language was a language other than English or French were less likely to experience ED visits (absolute risk reduction [ARR]=4.4%), hospitalisations (ARR=2.6%) or death (ARR=2.2%) when their family physician spoke a language that was mutually intelligible with their primary language.

How this study might affect research, practice or policy

  • Optimising the delivery of language-concordant primary care could result in significant decreases in the use of acute healthcare services and mortality at the population level.

  • The results of this study highlight the importance of systematically collecting the language(s) spoken by patients and healthcare providers in order to facilitate the matching of patients to healthcare providers who have proficiency in the patients’ primary language.

Introduction

Communication is an integral component of the patient–physician relationship. Physicians must be able to obtain a complete medical history to establish accurate and timely diagnoses, and then establish a management plan with their patient (including initiation of appropriate treatments). It is, therefore, not surprising that patients who face language barriers generally receive healthcare services of lower quality and safety.1–3 As immigration to Europe and North America continues to increase at an unprecedented rate,4 a growing proportion of the Western world’s population will experience language barriers as a result of living in a minority language situation (which occurs when an individual’s primary language does not correspond to the language spoken by the majority of the population in the region in which they live).5 This is especially true in Canada, where the number of residents living in a province that does not recognise their mother tongue as an official language increased from 9.5 million (27.3% of the population) in 2016 to 10.7 million (29.3% of the population) in 2021, reaching a record high proportion of the population since the first census was conducted in 1901.6 The rapid growth of Canada’s linguistic diversity is driven by an increase in the number of residents who primarily speak a language other than English or French (ie, allophones) while the proportion of residents who speak French (ie, francophones) has decreased.6

While there is consensus in the literature regarding health disparities across linguistic groups, there is limited research that assesses the impact of patient–physician language discordance, which occurs when patients and physicians do not have proficiency in a shared language.7–10 Studies in the USA have shown that Spanish-speaking patients with asthma and diabetes had better disease-specific outcomes when they received care from Spanish-speaking physicians.11–13 Another study conducted in Ontario, Canada showed that patients with tuberculosis (TB) who received care from respirologists who spoke their primary language were less likely to die within 1 year.14 Finally, two studies conducted in Ontario, Canada found that frail, older patients who were hospitalised and received language-concordant care had lower risks of adverse events,15 16 shorter stays in hospital16 and lower risks of in-hospital death.16 Several other studies have attempted to determine the impact of patient–physician language concordance/discordance by identifying patients receiving language-discordant care as those who were provided with an interpretation service7–10; such an approach combines two important factors, notably the patient’s language proficiency and the benefits of the interpretation service. Patient–physician language discordance warrants further investigation as it represents a potentially modifiable risk factor which could be the target of interventions aimed at improving the delivery of care to patients living in minority language situations (ie, by referring patients to physicians who have proficiency in their primary language or by adopting translation technology, whether in person or virtual).

Few studies have considered the impact of patient–physician language concordance on healthcare utilisation and overall survival in a population-based primary care setting. A study of Spanish-speaking patients with asthma found that patients who received language-concordant care had slightly fewer Emergency Department (ED) visits when compared with patients who received language-discordant care, although this difference did not reach statistical significance.11 Another study of patients with diabetes (who primarily spoke Haitian Creole, Portuguese and Spanish) found that patients were less likely to have diabetes-related ED visits when all of their primary care visits were with physicians who spoke their language.17 Lastly, a recent study of patients receiving primary care in California found that patients whose primary care provider spoke their language were more likely to see their primary care provider in follow-up and less likely to be admitted to hospital or see a specialist.18 Furthermore, prior studies of outpatients largely excluded frail, older patients, who are more likely to have communication problems and poor health outcomes when compared with the general population.19 20 The objective of this study was to assess the impact of patient–family physician language concordance on both hospital-based healthcare utilisation (ie, ED visits and hospitalisations) and overall survival among a cohort of home care recipients. We hypothesised that patients receiving language-concordant primary care would have better outcomes when compared with patients receiving language-discordant primary care.

Methods

Setting

Canada has a population of nearly 37 million people, of which 54.9% and 19.6% identify their mother tongue as English and French, respectively.6 Both English and French are recognised as official languages at the federal level while New Brunswick is the only province where both English and French have official language status in all branches of the provincial government.6 French is the official language in Quebec while English is the official language in the remaining provinces. However, several unilingual English provinces have enacted laws to protect the rights of francophones. For instance, in Ontario (Canada’s most populous province and the site of this study), the French Language Services Act identifies a number of government agencies (including healthcare facilities such as hospital and long-term care facilities) which are mandated by law to provide services in French in cities with at least 5000 francophones, and in regions of the province where francophones represent at least 10% of the population (generally Eastern and Northern Ontario). The French Language Services Act consists of 34 requirements, including (but not limited to) recruitment and hiring of staff with proficiency in French, adequate representation of francophones at all levels of management, systematic identification of patients’ linguistic identity and active offering of services in patients’ preferred language.21 In addition, many provinces, regional health authorities and hospitals have policies that mandate the offer of professional interpreter services to patients’ who are unable to receive care in their preferred language. Despite this, professional interpreters are not systematically employed across Canadian healthcare facilities,22 and the practice of using ad hoc or untrained interpreters remains pervasive within the Canadian healthcare system.23

Cohort description

We conducted a population-based retrospective cohort study in Ontario, Canada using administrative databases at ICES (formerly known as the Institute for Clinical Evaluative Sciences). Our baseline cohort consisted of all residents receiving publicly funded long-term home care services from 1 April 2010 to 31 March 2018. Residents were censored for 1 year or at death, whichever occurred first. We excluded residents who (1) were younger than 18 or older than 105 years at the index assessment, (2) were not eligible for the Ontario Health Insurance Plan (OHIP) on the date of the index assessment and (3) were enrolled in long-term care prior to their index assessment. Residents who completed multiple home care assessments were indexed at the time of their first home care assessment during the study period.

Patient and public involvement

As this study used routinely collected administrative data, patients were not directly involved in the design, recruiting, conduct, reporting or dissemination of this research.

Data sources

We used administrative databases at ICES, which is an independent, non-profit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyse healthcare and demographic data, without consent, for health system evaluation and improvement. We linked the Resident Assessment Instrument-Home Care (RAI-HC), which is a standardised data collection form for publicly funded home care assessments,24 to numerous administrative databases using unique encoded identifiers. The RAI-HC captures the baseline characteristics (including language) of Ontario residents receiving publicly funded home care services for at least 60 days.25 The RAI-HC also captures self-reported marital status, education and functional status. The Registered Persons Database (RPDB) provides Ontario residents’ date of birth, sex, postal code and date of death. Each resident’s postal code was then linked to Statistics Canada Census data to obtain a neighbourhood-level income quintile and urban/rural status of the residence.26 The Immigration, Refugees and Citizenship Canada (IRCC) Permanent Resident’s Database was used to identify immigrants who became permanent residents from 1985 to May 2017.27 We ascertained chronic conditions using algorithms validated by ICES and applied in previous studies (see online supplemental appendix 1). Finally, The National Ambulatory Care Reporting System (NACRS) and the Discharge Abstract Database (DAD) provided data on healthcare utilisation, including ambulatory care visits and inpatient hospitalisations.

The College of Physicians and Surgeons of Ontario (CPSO) database provides demographic information on all physicians in Ontario (including language(s) spoken by individual physicians) while the OHIP database records all physician billing claims. Finally, the Client Agency Programme Enrolment (CAPE) is a database, which identifies all patients who are rostered to a family physician. For patients who were not rostered to a family physician in the CAPE database, we assigned a family physician using OHIP billing claims, where we identified the individual family physician who submitted the largest dollar amount of OHIP-eligible billing claims for each patient in the last 2 years.28 29 Patients who were not rostered to a family physician in the CAPE database and who had not received care from at least one family physician in the 2 years preceding their index assessment were excluded from the cohort.

Exposure

We defined patient language from home care assessments (RAI-HC). During these assessments, interviewers are instructed to determine each resident’s primary language by listening to the language spoken in the resident’s household on their arrival and, if necessary, asking the resident and/or their family members to specify their primary language.24 We previously validated this language variable by comparing it to several self-reported languages obtained by the Canadian Community Health Survey (Batista et al, Institut du Savoir Montfort: unpublished data, 2020). We found that the level of agreement (determined with Cohen’s κ) was substantial for language most often spoken at home (κ=0.76). We considered three linguistic groups: anglophone (ie, primary language is English), francophone (ie, primary language is French) and allophone (ie, primary language is a language other than English or French). Residents who communicated with non-spoken languages (ie, artificial languages, sign languages) were excluded. Next, we obtained family physician language through the CPSO database. All physicians in Ontario are asked to submit the language(s) that they speak at the time of CPSO registration. We used this database to identify physicians who self-identified as speaking English, French and languages other than English or French.

After identifying both patient language and family physician language, we combined allophone languages to form groups of mutually intelligible languages (ie, languages that are different from one’s language but still readily understood).30–32 We chose to compare mutually intelligible languages, rather than individual languages, to account for the fact that patient language is recorded using either language family (Ibero-Romance) or individual languages corresponding to the language family (eg, Portuguese, Spanish). We retained the 10 most common groups of mutually intelligible languages in our cohort. A complete description of the classification of both patient language and family physician language is presented in online supplemental appendix 2. We defined primary care as language concordant if both the patient and their family physicians shared a mutually intelligible language, and we defined all other primary care as language discordant. Since all physicians in Ontario are required to speak English, all anglophone patients were considered to have received language-concordant primary care. For analyses which considered language concordance/discordance, we excluded patients whose language did not fall into the 10 most common groups of mutually intelligible languages (n=18 139, 3.6%) as well as patients whose family physician did not submit at least one language on their CPSO registration (n=267 523, 55.8%) for a final concordance analysis cohort of 211 565 patients.

Outcomes

We considered three outcomes: ED visits, hospitalisations and death occurring within 1 year of index assessment (ie, first home care assessment from 1 April 2010 to 31 March 2018). ED visits and hospitalisations were obtained from the NACRS and DAD, respectively, while the date of death was obtained from the RPDB. We recorded the time from study enrolment to the first occurrence of each outcome.

Statistical analysis

We performed descriptive analyses to compare patient characteristics and outcomes across linguistic groups, before and after stratifying by patient–physician language concordance/discordance. We used survival analysis using Cox regression to measure the impact of language concordance/discordance on the time to first ED visit, time to first hospitalisation and time to death. Survival times were defined as time to first ED visit, time to first hospitalisation and time to death. We fitted separate Cox models for francophones and allophones; patients receiving language-discordant care were the reference group in all analyses. When considering the outcomes of ED visits and hospitalisations, we used a cause-specific hazard function to account for the competing risk of death. Adjusted analyses included the potential confounders of age, sex, marital status, education, income quintile, geographical region, urban/rural residence, immigration status, number of chronic conditions, Charlson Comorbidity Index,33 Activities of Daily Living (ADL) scale,34 Instrumental Activities of Daily Living (IADL) scale,35 Cognitive Performance Scale36 and Changes in Health, End-Stage Disease, Signs and Symptoms score.37 We reported missing data in descriptive tables and included a level denoting missing data for categorical variables in all regression analyses. Statistical tests were two tailed and the significance threshold was set at 0.05.

Results

Baseline characteristics

A total of 497 227 home care recipients met eligibility criteria, including 398 819 anglophones, 11 050 francophones and 87 358 allophones. Among allophones in our cohort, the 10 most commonly spoken language groups were Italian (n=22 783, 26.1%), Mandarin (n=10 832, 12.4%), Ibero-Romance (n=7851, 9.0%), Indo-Aryan (n=7268, 8.3%), East Slavic (n=4151, 4.8%), West Slavic (n=3546, 4.1%), Dravidian (n=3355, 3.8%), West Germanic (n=3350, 3.8%), Greek (n=3074, 3.5%) and Arabic (n=3009, 3.4%). The remaining allophones (n=18 139, 20.8%) were included in descriptive statistics but excluded from subsequent analyses which considered patient–physician language concordance/discordance.

The baseline characteristics of the cohort are presented in table 1. Most home care recipients were anglophone (80.2%) while francophones (2.2%) and allophones (17.6%) represented a minority of the cohort. Francophones and allophones were slightly older than Anglophones. When compared with anglophones, a smaller proportion of francophones and allophones had completed high school or university, while a greater proportion of francophones and allophones resided in lower-income neighbourhoods. Francophones tended to live in rural areas (especially in Eastern and Northern Ontario) while allophones lived almost exclusively in urban areas. The proportion of residents who immigrated to Canada was greatest for allophones.

Table 1
|
Baseline characteristics of home care recipients (n=497 227)

The health characteristics and functional status of the cohort are presented in table 2. The overall burden of disease (characterised by the number of chronic conditions and the Charlson Comorbidity Index) was similar across linguistic groups. The proportion of residents who required assistance with ADLs and IADLs was greater for allophones when compared with both anglophones and francophones, who had similar functional limitations. Finally, both francophones and allophones had more cognitive impairment than anglophones.

Table 2
|
Health characteristics and functional status of home care recipients (n=497 227)

A total of 12 973 family physicians provided care to the 497 227 home care recipients included in our cohort. Only 7.9% of the family physicians in our study reported speaking French while 21.6% of family physicians reported speaking a language other than English or French. Language was missing for 5289 physicians (40.8%). After excluding home care recipients whose family physician was missing data on language, our final cohort consisted of 5510 francophone and 27 818 allophone home care recipients (see figure 1). More than half of francophone patients (61.3%) received primary care from French-speaking family physicians while relatively fewer allophone patients (43.9%) received primary care in a language that was mutually intelligible with their primary language.

Figure 1
Figure 1

Study flow diagram. OHIP, Ontario Health Insurance Plan; RAI-HC, Resident Assessment Instrument-Home Care.

Outcomes

As shown in table 3, a similar proportion of anglophones and francophones had ED visits (62.1% vs 63.3%, respectively) and hospitalisations (42.2% and 42.2%, respectively) during the study period while allophones had significantly fewer ED visits (55.7%) and hospitalisations (37.0%). The proportion of allophones who died during the study period (17.0%) was also lower when compared with both anglophones (21.5%) and francophones (20.9%).

Table 3
|
Unadjusted outcomes for all home care recipients, stratified by patient language (n=497 227)

Allophones who received language-concordant primary care experienced fewer ED visits (53.1% vs 57.5%; p<0.01) and hospitalisations (35.0% vs 37.6%; p<0.01) when compared with those who received language-discordant primary care. Furthermore, the proportion of allophones who died within 1 year of index assessment was significantly lower for those who received language-concordant primary care when compared with those who received language-discordant primary care (14.4% vs 16.6%; p<0.01). The proportion of francophones who had at least 1 ED visit during the follow-up period was slightly lower among those who received language-concordant primary care compared with those who received language-discordant primary care (61.6% vs 64.4; p=0.04). The occurrence of hospitalisations and death among francophones was not impacted by the language of their family physician (see table 4).

Table 4
|
Unadjusted outcomes for francophone and allophone home care recipients, stratified by primary care language concordance/discordance (n=33 328)

In the adjusted regression analyses, allophones who received language-concordant primary care had lower risks of ED visit (adjusted hazard ratio [HR] 0.91, 95% CI 0.88 to 0.94), hospitalisation (aHR 0.94, 95% CI 0.90 to 0.98) and death (aHR 0.87, 95% CI 0.82 to 0.93) than allophones who received language-discordant primary care (figure 2). There was no statistically significant difference in the risks of ED visits, hospitalisations and death when comparing francophones who received language-discordant primary care to those who received language-concordant primary care (figure 2).

Figure 2
Figure 2

Adjusted outcomes for francophone and allophone home care recipients stratified by language concordance/discordance of primary care. Values to the left of the line of null effect denote lower risk of ED visit, hospitalisation and death among patients receiving language-concordant primary care; values to the right of the line of null effect denote higher risk of ED visit, hospitalisation and death among patients receiving language-discordant primary care. Hazard ratios adjusted for age, sex, marital status, education, income quintile, geographical region, urban/rural residence, immigration status, number of chronic conditions, Charlson Comorbidity Index, ADL scale, IADL scale, cognitive performance scale and Changes in Health, End-Stage Disease, Signs and Symptoms score. †Statistical significance at the 0.05 level. ADL, Activities of Daily Living; ED, Emergency Department; IADL, Instrumental Activities of Daily Living.

Discussion

In this study of home care recipients receiving primary care from family physicians in Ontario, Canada, we found that the proportion of ED visits, hospitalisations and death during the 1-year study period was lowest for those who spoke a language other than English or French (ie, allophones). These findings are consistent with those of previous studies, which have found that allophones living in minority linguistic communities are less frequent users of hospital-based healthcare services.38 39 This has generally been attributed to barriers to accessing care7 40 41 and to cultural differences resulting in a preference for community-based care rather than hospital-based care.42 43 To our knowledge, only one study has compared mortality rates across all linguistic groups in the Canadian population; this study found that allophones living in Quebec had significantly lower all-cause mortality than both anglophones and francophones.44 Differential mortality among allophones may also partly be explained by the healthy immigrant effect, which postulates that individuals who have immigrated to Canada have more positive health status on arrival when compared with age-matched and sex-matched Canadian-born citizens due to lifestyle behaviours and self-selection related to immigration policies and procedures.45 The advantage in health status seen among Canadian immigrants has been shown to dissipate after 2–10 years45; however, differential mortality persists, with one study reporting lower age-standardised mortality rates at 20 years after immigration to Canada.46 Only 1 in 10 allophones in our study were identified as immigrants; however, our study was limited by the fact that we were only able to identify recent immigrants (ie, those who immigrated since 1985), which likely resulted in misclassification of immigration status for older home care recipients who may have immigrated before 1985 (although there is expected waning of healthy immigrant effect in these individuals).

Allophones who received language-concordant primary care were less likely to use hospital-based healthcare services and to experience death when compared with allophones who received language-discordant primary care. We hypothesise that these results could be due, at least in part, to language barriers contributing to lower levels of satisfaction with their family physician as well as poorer quality of care received.1 7 40 41 Clear and effective patient–physician communication is associated with enhanced patient cooperation and engagement,47 which in turn has the potential to improve patient outcomes. These findings are supported by American studies which have shown that Spanish-speaking patients with asthma who received care from Spanish-speaking physicians were less likely to miss appointments, omit medications or visit the ED.11 Furthermore, Spanish-speaking patients with diabetes who receive language-discordant primary care had worse glycaemic control12 and low-density-lipoprotein control,13 as well as poorer participation in diabetic foot care programmes.48 Another study conducted in Ontario, Canada showed that patients with TB who received care from respirologists who spoke their primary language were less likely to die within 1 year.14 Poor patient–physician communication can affect the ability to obtain accurate information to be able to make timely diagnoses and formulate management plans,1 3 41 which in turn could result in patients seeking hospital-based healthcare services to accelerate their diagnostic workup or to address unmet needs (including lack of information due to communication barrier).

Interestingly, outcomes for francophones were not statistically impacted by the language of their family physician. This finding can likely be attributed to the relatively higher rate of bilingualism among francophones when compared with allophones. According to the 2021 Census, the proportion of francophones and allophones (defined by language spoken most often at home) over the age of 65 in Ontario, Canada who are able to conduct a conversation in English is 86.6% and 68.7%, respectively.49 Furthermore, a previous study of home care recipients in Ontario, Canada found that more than half of allophones had low English proficiency, while less than 10% of francophones were deemed to have low English proficiency.50 Therefore, relatively fewer francophones will experience language barriers when interacting with their family physician (all of whom should be able to speak English since this is a requirement to practice medicine in Ontario), and the language barriers experienced by francophones are likely to be less severe when compared with those experienced by allophones.3 Finally, while French is not recognised as an official language at the provincial level, the French Language Services Act identifies a number of healthcare facilities (both community based and hospital based) which are required by law to provide services in French in regions of the province with the highest proportion of francophone residents (defined as geographical regions where francophones make up at least 10% of the population).21 This includes, but is not limited to, public health programmes providing support for addictions, dementia, disability, mental health, preventative care and sexually transmitted infections.21 As a result, francophones may have been more likely to access language-concordant community-based resources rather than use hospital-based healthcare services.

The findings of this study contribute to the paucity of research on the topic of patient–provider language concordance/discordance in Canada. We previously showed that home care recipients admitted to a hospital in Ontario who received more than half of their care from physicians who shared a mutually intelligible language had lower risks of adverse events, shorter stays in hospital and lower risks of in-hospital death.16 By considering patients in two different settings (ie, inpatient vs outpatient), this study provides additional information regarding the way in which linguistically concordant/discordant care is offered to patients in Ontario. First, this study determined language concordance by using the language(s) of the patient’s primary care provider while our prior study determined language concordance by calculating a weighted average of the languages spoken by all physicians who provided care during the hospitalisations.16 Next, the outcomes under study were different, with an emphasis on hospital-based healthcare utilisation (ED visits, hospitalisations) in this study, versus in-hospital outcomes (harm and length of stay) in our prior study.16 While less than half (43.9%) of allophones received language-concordant care in this study (outpatient setting), only 1.6% of allophones received language-concordant care in our prior study (inpatient setting).16 A similar trend was observed for francophones, who were more likely to receive language-concordant care as outpatients (61.3%) rather than as inpatients (44.4%).16 Our prior study also identified much larger effect sizes for inpatients when compared with the effect sizes identified for outpatients in this study,16 which may be explained by access to and availability of language-concordant care (greater in the outpatient setting compared with the inpatient setting, where patients may not have the opportunity to seek care from a physician who speaks their primary language).

Limitations

We obtained patient language from home care assessments, during which interviewers recorded the patient’s primary language.24 Since interviewers can only record one language for each home care recipient, we may have overestimated the prevalence of language discordance among those who speak multiple languages (including English). However, such misclassification would result in a bias towards the null, as patients would be incorrectly identified as having received language-discordant care (higher risk of poor outcomes) when they in fact received language-concordant care (lower risk of poor outcomes). Furthermore, it is possible that interviewers may have recorded home care recipient’s language as English to avoid the need to obtain interpreter services. This would result in francophones and allophones being incorrectly identified as Anglophones, which would also minimise the disparities experienced by francophones and allophones when compared with anglophones. Next, we do not know whether francophone and allophone patients identified as receiving language-concordant care did in fact communicate with their physician in their primary language or, in the case of allophones included in a group with multiple languages, whether the communication occurred in a language that was mutually intelligible with their primary language. As above, misclassification of allophone patients as having received language-concordant care when they in fact received language-discordant care (eg, because they experienced a language barrier when interacting with a physician who had proficiency in a language that was mutually intelligible but not identical to their primary language) should bias the results towards the null. We also do not know whether interpretation services were used for patients identified as receiving language-discordant care. It is likely that some francophone and allophone patients had family members, friends or other healthcare providers who provided interpretation during their interactions with their family physician. Unfortunately, professional interpreters are not systematically employed across Canadian healthcare facilities,22 and studies have shown that non-professional interpreters (ie, family members, friends and hospital employees) do not consistently improve outcomes. Therefore, we feel that interpreter use is unlikely to have had a significant impact on the outcomes reported in our study.51 52 Finally, physician language obtained from the CPSO has not yet been validated and, unfortunately, was missing for almost half of physicians in our study. However, since physician language is recorded prior to and independently from patient outcomes, any misclassification should be non-differential, thereby leading the results to be biased towards the null.

Conclusions

Patients whose primary language was a language other than English or French were less likely to use hospital-based healthcare services or to experience death when they received language-concordant primary care from family physicians. The results of this study highlight the importance of systematically collecting the language(s) spoken by patients and healthcare providers, to allow healthcare systems to implement strategies to facilitate the matching of patients to healthcare providers who have proficiency in their primary language, as well as access to interpretation services when patient–physician matching is not possible.