Distance to primary care and its association with health care use and quality of care in Ontario: a cross-sectional study.

Methods

Setting and context
The study setting was Ontario, which has a population of almost 16 million,20 where all permanent residents have financial coverage for family physician care under the Ontario Health Insurance Plan (OHIP). Despite this publicly funded health insurance, about 2.3 million Ontarians lack access to a regular source of primary care.21
The primary care model in Ontario has evolved from a fee-for-service system of independent physicians to group-based practices based on patient enrolment.22 More than 75% of Ontario’s population is enrolled with a family physician. A patient enrolment model creates a formal relationship known as rostering, attaching the patient to a primary care provider. These arrangements involve a dual commitment from the patient and physician, which can provide benefits such as enhanced access and quality of care, as well as stronger patient–physician relationships.23

Study design and population
We conducted a population-based cross-sectional study using administrative data of all patients enrolled with a family physician in Ontario as of Mar. 31, 2023, and living in an urban or suburban region of the province (Rurality Index for Ontario [RIO] score of 0 to 39).24 We included only patients enrolled with a family physician in a patient enrolment model to ensure that patients had a regular source of care and that we were examining comparable populations. We excluded patients without a valid health card, non-Ontario residents, those with birth dates after the index date or death dates before the index date, those missing data on birth date or sex, patients not in a patient enrolment model, patients in a patient enrolment model but not rostered (i.e., those who could see any physician in the group but who did not have an assigned family physician), patients who were virtually rostered (i.e., patients who were linked to a physician based on billing history but not formally enrolled),25 and patients who had visited a community health centre for primary care in the previous 2 years.

Data sources
We identified the study population using the April 2023 Primary Care Population (PCPOP) data set, an ICES-derived database that captures all people in Ontario who are alive and eligible for OHIP coverage. ICES is an independent, nonprofit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze health care and demographic data, without consent, for health system evaluation and improvement. The PCPOP database includes information on patient age, sex, primary care rostering, new migrants to Ontario (i.e., registered in OHIP in the previous 10 years), use of health services (e.g., emergency department visits, primary care visits), and indicators of primary care such as cancer screening. The data were linked with other databases, including the Registered Persons Database (RPDB) for patient postal codes and demographics and the Corporate Provider Database (CPDB) for physician office postal codes. Postal codes were used to derive the RIO score and neighbourhood income quintile of patients. Data sets were linked using unique encoded identifiers and analyzed at ICES. Additional information on the data sources used can be found in Appendix 1, available at www.cmaj.ca/lookup/doi/10.1503/cmaj.250265/tab-related-content.

Distance to primary care
The primary exposure variable was the patient’s travel distance to their family physician. We calculated distances using the difference in latitudinal and longitudinal coordinates between patients’ postal codes and those of their family physicians from Statistics Canada’s Postal Code Conversion File. Straight-line distances were categorized as 10 km or less, 11 to 30 km, 31 to 50 km, 51 to 100 km, 101 to 150 km, 151 to 200 km, or more than 200 km.
We also took a 10% random sample of patients and estimated driving distance to family physicians along Ontario’s road network and drive times (based on posted speed limits) to assess for congruence with straight-line distances. Drive times were categorized as 15 minutes or less, 16 to 30 minutes, 31 to 45 minutes, 46 to 60 minutes, 61 to 90 minutes, 91 to 120 minutes, or more than 120 minutes.

Health care use and quality of care
The primary outcomes were access to care and quality of patient care relative to the index date. We assessed emergency department use over the preceding year from the index date and identified any emergency department visits and any nonurgent emergency department visits based on the Canadian Triage and Acuity Scale. We assessed primary care utilization over the preceding 2 years from the index date and included primary care visits (the number of core primary care visits), and continuity of care (i.e., mean continuity of care and the number of patients with 3 or more visits to their own rostering physician or group). Core primary care visits included visits to any family physician — not necessarily their own family physician or rostering group — where the physician billed at least 1 of the core primary care fee codes (Appendix 1). We calculated continuity of care as the number of visits to the rostering physician or rostering group, divided by all primary care visits to all family physicians. We used a 2-year look-back for primary care visits and a 1-year look-back for emergency department visits to reflect their differing frequencies of use.
Other quality of care outcomes included colorectal cancer screening in the previous 2 years (fecal immunochemical tests) or 10 years (colonoscopy), breast cancer screening (mammogram) in the previous 2 years, and cervical cancer screening (Pap test) in the 3 years before the index date. Females eligible for breast cancer screening were defined as those aged 52 to 69 years, excluding those with a history of breast cancer or mastectomy. Females eligible for cervical cancer screening were defined as those aged 21 to 69 years, excluding those with a history of cervical cancer or hysterectomy. People eligible for colorectal cancer screening were defined as those aged 52 to 74 years, excluding those who had a history of colorectal cancer or inflammatory bowel diseases.26
A full list of variables, their definitions and data sources can be found in Appendix 1.

Statistical analysis
We completed analyses using SAS, Epi Info 7.2.6.0 (US Centers for Disease Control and Prevention), as well as R 4.4.2 and R Studio 2024.09.0. We derived travel distances and travel times using ArcGIS. Patient characteristics, quality of care indicators, and distance and time variables were cross-tabulated for descriptive analyses. We completed straight-line distance analyses for the entire study population, and we conducted a sensitivity analysis of drive distances and times for 10% of the total study population. The 10% sample was randomly selected from each dissemination area27 to ensure representation from a small, relatively homogeneous population that reflected the area’s total population.
We used descriptive statistics, including proportions and percentages. We used the χ2 test for trend to assess associations between distance and emergency department visits, no core primary care visits, continuity of care to rostering group and family physician, and cancer screening (breast, colorectal, and cervical). We performed weighted linear regression to assess trends in mean continuity of care scores to the patient’s rostering group and physician across distance and time categories, with weights applied to account for unequal group sizes.
Using previous studies, we also collapsed distance and drive time into dichotomous categories (≤ 30 km v. > 30 km and ≤ 30 min v. ≥ 30 min) to further assess associations between population characteristics, access to care, continuity of care, and cancer screening.28–31 We conducted a sensitivity analysis using dichotomous distance cut-offs of 50 km (≤ 50 km v. > 50 km).32–34 To assess associations between 2 independent categorical variables, we used cross-tabulation and χ2 tests of independence. We calculated p values, crude odds ratios (ORs) and their 95% confidence intervals (CIs). For binary comparisons of mean continuity scores, we used weighted analysis of variance to compare differences in means across dichotomous distance and time categories, incorporating weights to reflect varying sample sizes between groups.

Ethics approval
The use of data in this study was authorized under section 45 of Ontario’s Personal Health Information Protection Act, 2004 and did not require research ethics board approval.

Results
We included 9 967 955 patients in Ontario in our straight-line analysis after excluding those who were not in a patient enrolment model (n = 2 967 794), those who were part of a patient enrolment model but not rostered to a family physician or were virtually rostered (n = 1 500 426), those attached to community health centres (n =174 168), and those who were missing data (n = 7224) (Figure 1). Table 1 summarizes the characteristics of the population, including age, sex, income quintile, and newcomer status. There were 1 261 112 (12.6%) patients who lived more than 30 km from their family physician.
Those who lived farther than 30 km from their family physician were more likely to be male (OR 1.08, 95% CI 1.07 to 1.08), younger than 65 years (OR 1.55, 95% CI 1.54 to 1.55), live in a neighbourhood of the lowest income quintile (OR 1.04, 95% CI 1.03 to 1.04), and be newcomers to Ontario (OR 1.06, 95% CI 1.05 to 1.06) (Appendix 2, Table S2.1 available at www.cmaj.ca/lookup/doi/10.1503/cmaj.250265/tab-related-content).

Health care utilization
Patients who lived farther from their family physician had more emergency department visits overall and for nonurgent concerns (Figure 2). Increasing distance was associated with higher odds of an emergency department visit, with the OR peaking at 1.30 for patients living 151 to 200 km away, compared with those living closer than 10 km to their family physician (p < 0.001) (Appendix 2, Table S2.2).
To assess primary care utilization, we looked at core primary care visits (Figure 3). The χ2 test for trend highlighted that as distance increased, so did the proportion of patients with no core primary care visits to those with any visits (Appendix 2, Table S2.3).

Quality of care
We assessed quality of care by measuring continuity of care and preventive screening rates for patients based on their distance from their rostering family physician.
We calculated mean continuity of care scores for visits to patients’ rostering physician group and to their own rostering family physician (Figure 4). For the former, mean continuity of care scores showed marked declines by distance, from 69.6% for patients who lived less than 10 km away to 42.2% for patients who lived greater than 200 km from their family physician. Similarly, when we considered visits to patients’ own family physicians only, mean scores declined by distance, from 63.7% for patients who lived less than 10 km away to 38.0% for patients who lived greater than 200 km from their family physician. Weighted linear regression confirmed that as straight-line distance increased, the mean continuity of care scores to the rostering group (p < 0.001) or rostering physician decreased (p < 0.001) (Appendix 2, Table S2.4).
We measured preventive screening rates for patients based on their distance from their rostering family physician (Figure 5). The χ2 test for trend showed that, in general, the farther patients lived from their rostered family physician, the less likely they were to have colorectal cancer screening with either fecal immunochemical test (in the past 2 years) or colonoscopy (in the past 10 years), mammogram in the previous 2 years, or cervical cancer screening with a Pap smear in the last 3 years (Appendix 2, Table S2.5).

Outcomes with a 30-km threshold
Compared with patients who lived 0 to 30 km from their family physicians, those who lived farther than 30 km away had greater odds of emergency department use, having no visits to their family physician, and not being screened for cancer (Table 2). In addition, patients who lived farther than 30 km from their family physician had a lower average continuity of care scores to their own group and physician than those who lived within 30 km (Table 3).

Sensitivity analyses
Results from the sensitivity analyses showed associations were similar when using a dichotomous distance cut-off of 50 km (0 to 50 km v. > 50 km), with ORs generally higher than when we used the 30-km cut-off for emergency department visits, core primary care visits, cancer screening, and continuity of care (Appendix 2, Table S2.6 and Table 2.7). We also found that results from analyses using drive distances (Appendix 3, available at www.cmaj.ca/lookup/doi/10.1503/cmaj.250265/tab-related-content) and drive times were consistent with straight-line distance analyses (Appendix 4, available at www.cmaj.ca/lookup/doi/10.1503/cmaj.250265/tab-related-content). The Pearson correlation between our 100% straight-line distance sample and 10% drive distance sample was 0.91. The Pearson correlation between the 100% straight-line distance sample and the 10% drive time sample was 0.93.

Interpretation
Many people in Canada lack access to primary care. Our findings highlight that even those with a family physician often travel long distances to access care. We found that more than 1 in 10 patients in urban or suburban regions who were rostered to a family physician in a patient enrolment model in Ontario lived farther than 30 km from their family doctor. These patients had a higher odds of being male, being younger than 65 years, having lower socioeconomic status, and being newcomers to Ontario. They also had higher odds of visiting an emergency department, not seeing any family physician, not having continuity of care with their family physician or team, and not being screened for breast, colon, or cervical cancer.
Countries with strong primary care tend to have patients live close to their providers.35 In the Netherlands, almost all citizens have access to a general practitioner. Although people can choose their provider, enrolment at a clinic can be rejected if the distance between the clinic and the patient is high.17 On average, patients in the Netherlands live 1 km from the closest general practice.18 Similarly, a study done in Norway found that travel time from home to someone’s general practitioner was short for most of the population (< 5 min at median and < 20 min at 90th percentile).19 This study also found that residents move closer to their general practitioners when they start needing services, especially patients who used to live far away.
In a 2016 systematic review, several studies showed a distance decay association, whereby greater distances to health care services are associated with reduced access to health care services and poorer health outcomes.36 One study found that longer travel times between patients and their general practitioner significantly increased the risk of being diagnosed with breast or colorectal cancer at a later stage.37 Two other studies found that patients with diabetes who had longer drive distances to their primary care facility had poorer glycemic control.38,39 In this review, none of the studies looked at the rate of emergency department use, primary care use, continuity of care, or cancer screening rates based on the distance between patients and their own primary care provider.
Our findings of lower continuity of care with greater distance are in keeping with other research studies that found patients who lived farther from their family physician were more likely to visit a walk-in clinic. Rahman and colleagues8 showed that almost 1 in 5 patients visit walk-in clinics to receive care given the clinics’ closer proximity than their family physicians. Lapointe-Shaw and colleagues40 similarly found that patients who were enrolled with a family physician were more likely to visit walk-in clinics closer to their homes than the enrolling family physicians’ offices. These trends disrupt continuity of care and have negative implications. Continuity of care improves the quality of care for patients and is cost-effective.41–43 Studies have found that continuity of care was associated with fewer hospital visits in the United States44 and ambulatory care–related hospitalizations in Canada.45 A large-scale cohort study in the Netherlands also found that low continuity of care in primary care was associated with a higher risk of death.46 An important finding from our study highlighted how it is not just lack of a family physician47 or appointment availability that drive emergency department use48 but also that the distance to one’s family physician affects access to primary care.
Although we have shown that distance affects health care utilization and access to cancer screening at a population level, we also recognize that there are circumstances where individuals may choose to travel past their nearest provider to receive treatment. 49–51 A scoping review explored how patients chose their physicians for continuous outpatient care and why they were willing to bypass the nearest physician. The review found that factors related to the individual patient (e.g., personal experiences, location of workplace, recommendations from family and friends, availability of a car), quality of care (e.g., doctor–patient relationship, trust, range of services, organization of practice), and access (e.g., degree of urbanity, community fit, waiting time for appointments, availability of public transport) were considered by patients.52 However, patients who prefer to receive care closer to home should have the option to do so.
A common indicator of access to primary care is the proportion of patients who have a family physician or primary care provider.53 However, our study showed gaps in quality of care and health care use that arise when patients live far from their family physician. More research is needed to understand the proportion of patients who are seeking a new family physician because theirs is too far away. Incorporating distance to a family physician can provide policy-makers with a more nuanced understanding of unmet primary care demand. Our findings suggest that reforms should prioritize offering primary care a minimum of 30 km from a patient’s home.

Limitations
We studied only patients attached to family physicians in Ontario’s patient enrolment models. Although this is a limitation, it allowed us to compare populations of patients who typically have similar access to care based on their model of care. Patient–physician distance was sourced from the RPDB and CPDB databases; inaccuracies in addresses may exist because of changes in residence or practice location and data entry errors. In this study, we used straight-line distances as our primary distance measure, given our large sample size (nearly 10 million people) and related computational limitations in measuring drive time and drive distances for this large sample. The literature supports using straight-line distances as a proxy for drive time.54–56 To address this limitation, we conducted sensitivity analyses with drive time and drive distances for a 10% sample (nearly 1 million people), which showed high correlations between the 2 methods. We included patients younger than 18 years, recognizing that distance affects their care, but did not include family-level data, given their unavailability in our data sets. The lack of household linkage may have led to an overestimation of some outcomes owing to unmeasured shared behaviours among members of the same household. We acknowledge the limitations of the CTAS scores in identifying primary care–sensitive conditions; future work will incorporate additional appropriateness measures and examine how after-hours primary care access affects emergency department use. We used 4 demographic variables — age, sex, income quintile, and newcomer status — based on the availability of data. We acknowledge that other important factors (e.g., education, language, race or ethnicity, Indigenous identity, employment, occupation, disability status, transportation access, household composition, social support) may also influence proximity to family physicians and health-seeking behaviour. These limitations underscore the importance of ongoing efforts to collect comprehensive sociodemographic data for population-level research.57 Although we did not adjust outcomes to demographic factors because of data and time constraints, we explored their influence based on distance, noting a dose–response relationship as outcomes worsened with increasing distance. However, as this study was cross-sectional, causal inferences should be made with caution.

Conclusion
Among patients living in an urban or suburban region who were rostered to a family physician in a patient enrolment model in Ontario, more than 1 in 10 lived farther than 30 km from their family physician. We found that having access to primary care close to home was associated with higher continuity of care, lower emergency department use, and more recommended cancer screening. Primary care reforms should prioritize providing people access to care close to home.

Supplementary Information