Geospatial access predicts stage at presentation and outcomes for breast cancer patients in South West Nigeria: A population-based study

Abstract

Background:

The majority of women in Nigeria present with advanced stage breast cancer. To address the role of geospatial access, we constructed a geographic information system-based model to evaluate the relationship between modeled travel time, stage at presentation and overall survival among breast cancer patients in Nigeria.

Methods:

Consecutive patients were identified from a single-institution, prospective breast cancer database (May 2009-January 2019). Patients were geolocated and travel time to the hospital was generated using a cost-distance model that utilized open source data. The relationships between travel time, stage at presentation and overall survival were evaluated with logistic regression and survival analyses. Models were adjusted for age, level of education and socioeconomic status.

Results:

From 635 patients, 609 were successfully geolocated. The median age of the cohort was 49 years (IQR 40–58) and 84% presented with ≥ stage 3 disease. Overall, 46.5% underwent surgery and 70.8% received systemic chemotherapy. The median estimated travel time for the cohort was 45-minutes (IQR 7.9–79.3). Patients in the highest travel time quintile had a 2.8-fold increase in the odds of presenting with stage 3 or 4 disease relative to patients in the lowest travel time quintile (p=0.006). Travel time ≥30 minutes was associated with an increased risk of death (HR 1.65, p=0.004).

Conclusion:

Geospatial access to a tertiary care facility is independently associated with stage at presentation and overall survival among breast cancer patients in Nigeria. Addressing disparities in access will be essential to ensure the development of equitable health policy.

PRECIS STATEMENT:

Geospatial access to cancer care is associated with stage at presentation and overall survival among breast cancer patients in Nigeria. As the country’s breast cancer control system matures, measuring access will be essential to ensure the development of equitable health policy.

Introduction:

The incidence of breast cancer is increasing across sub-Saharan Africa¹. In Nigeria, the number of new cases has increased by over 80% in the last four decades, making breast cancer the most common cause of cancer-related mortality with an age-standardized incidence of 52–63/100,000^1–3. Unfortunately, delays in presentation are common in Nigeria. This results in advanced stage diagnosis and lower survival relative to those treated in high-income countries^4–8.

The etiology of delayed presentation in Nigeria is multifactorial and includes a lack of awareness of breast cancer symptoms and treatment options as well as inadequate access to healthcare services^5,9–11. Access to healthcare may negatively affect healthcare utilization through patient factors (e.g. financial constraint) or through the organization of health services organization (e.g. equipment availability)¹². Geographic barriers to access are often exaggerated in lower-resource countries such as Nigeria, where >50% of the population is rural and the transportation infrastructure is poor^13,14. Indeed, travel time from place of residence to treatment facility is inversely related to receipt of treatment, prognosis, and quality of life for patients with a variety of cancers¹⁵. This may be particularly true for contemporary, multimodality breast cancer management, which requires numerous visits to a healthcare facility over an extended period of time.

In sub-Sahara Africa, there is a paucity of data on the relationship between geospatial access to tertiary-level care and breast-cancer specific outcomes¹⁵. To date, studies have utilized patient reported travel times analyzed as an ordinal variable and straight-line distance measurement, which provide an imprecise and potentially biased proxy for an access-outcomes analysis^10,16,17. As the implications for integrated cancer care policy are significant, we utilized a robust geographic information system (GIS) based model to analyze the relationship between geospatial access, stage of breast cancer at presentation and survival at a tertiary referral center in southwestern Nigeria.

Methods:

Registry Data

Consecutive patients were identified from a single institution, prospective breast cancer database. Patients without histologically confirmed invasive breast cancer (e.g. invasive ductal carcinoma) and patients primarily treated at another facility were excluded. Prior to 2016, data was collected retrospectively. The database is maintained by the African Research Group for Oncology (ARGO) at Obafemi Awolowo University Teaching Hospital (OAUTH) and captures data on sociodemographics, clinical presentation, histopathology, treatment (i.e. surgery, chemotherapy, radiotherapy), and long-term outcomes. Patient information is gathered at several time-points, including initial consultation, time of operation and during routine follow-up, by trained research assistants. The staging system used in this database reflects the American Joint Committee on Cancer 8^th edition. The metastatic work-up included a chest x-ray and abdominal ultrasound. Systemic therapy was administered by the treating surgeon. Adjuncts to surgery (e.g. systemic therapy) were offered to all patients as per national and international guidelines within the constraints of the available resources. Overall Survival (OS) was calculated from date of presentation until death and patients alive at last follow up were censored.

Environment

All of the patients captured in the breast cancer database were treated at OAUTH, in Ile-Ife, Osun State. This facility is the principal referral hospital for cancer care for the state (i.e. Osun State), as well as the adjacent states of Ekiti and Ondo. The geographic catchment area for OAUTH was defined by the borders of these three states (i.e. Osun, Ondo, and Ekiti), with a population of >12 million. In Osun State, OAUTH is the surgical referral center for six State Hospitals and five General Hospitals^18,19.

Geospatial Data

The primary treatment facility (i.e. OAUTH) as well as the residential locations of the presenting patients were geolocated using Google Earth™ (Google, Mountain View, CA). Where place of residence could not be geolocated based on information provided, the nearest bus stop, post office or city-center were used, respectively. A georeferenced national road network was obtained from OpenStreetMap (OpenStreetMap Foundation, Cambridge, United Kingdom). Each road was classified as primary (national highways), secondary (intra-state roads), and tertiary (minor roads, unpaved roads). All road data was cleaned and topographically verified using ArcMAP v10.2 (Esri, Redlands, CA). Road speed data were based on national traffic laws whereby primary, secondary, and tertiary roads were assigned speed limits of 80, 50 and 30 kmh⁻¹, respectively²⁰. Road speed weighting was verified by one of the authors (MOO).

Measuring Access

Travel time estimates for each patient to OAUTH were generated using a cost distance analysis as previously described^20–22. This method calculates the cumulative travel time associated with travel from any point in the study area to the treatment facility using the most time-efficient route over the national road network. Any portion of a route that included travel over terrain without roads was assigned a walking speed of 5 kmh⁻¹. By overlaying the residential locations of the patient cohort on this output, the estimated travel times from this location to OAUTH were extracted and incorporated into the subsequent statistical models. The geographic catchment area around OAUTH (i.e. Osun, Ondo and Ekiti) was used as a sensitivity analysis to evaluate the impact of referral bias (i.e. more advanced disease at a tertiary care facility) and income mobility (i.e. patients seeking care at preferred vs. nearest facility).

Statistical Analysis

The relationship between estimated travel time and advanced stage of presentation was estimated using logistic regression modelling. Patients without any residential location data were excluded. Odds ratios were adjusted for age, level of education and low socioeconomic status (SES). A three-tier SES classification was based on self-reported occupation, which has been previously validated in a local population²³. The relationship between death and geospatial access was examined using a cox proportional hazards model and Kaplan-Meier survival curves that compared patients living within 30 minutes from OAUTH to those living ≥30 minutes from OAUTH. Chi square tests, student t-tests or likelihood ratio tests were used, where appropriate. A significant result was defined as p < 0.05. All statistical analyses were performed using Stata v14 (StataCorp, College Station, TX).

Results

Between May 2009 – January 2019, 635 patients presented to OAUTH for breast cancer treatment. Location data was available for 609 patients, representing 95.9% of the cohort. The precise place of residence was located for 293 patients (48.1%). For 28.9%, 18.2% and 4.8% of patients, the nearest bus stop, post office or city center was used, respectively. The median age was 49 years (IQR 40–58) and the majority of patients were of lower SES and had at least primary education. Eighty-four percent of patients presented with advanced disease (i.e. stage 3–4). Only 21.7% of patients had immunohistochemistry for estrogen (i.e. ER) and human-epidermal growth factor receptor (i.e. HER-2) status. Overall, 46.5% underwent surgery and 70.8% received systemic chemotherapy. At a median follow-up of 12 months (IQR 3–27 months), overall survival was 66.8%. Complete baseline characteristics for the study cohort are presented in Table 1. The median estimated travel time for the cohort was 45-minutes (IQR 7.9–79.3). National travel times to OAUTH as well as the geographic distribution of the study patients are depicted in Figure 1.

Table 1:

Baseline characteristics of study cohort

	No. (%) or Median (IQR)
Total	609
Age (years)	49 (40–58)
Gender
Female	598 (98.2)
Stage
1	8 (1.4)
2	88 (15.0)
3	355 (60.6)
4	135 (23.0)
Travel Time	45.0 (7.9–79.3)
Received Surgery	283 (46.5)
Received Chemotherapy	431 (70.8)
Received Radiotherapy	51 (8.5)
Socioeconomic status
Low	417 (68.5)
Middle	185 (30.4)
High	7 (1.2)
Level of Education
None	56 (9.3)
Primary	196 (32.4)
Secondary	141 (23.3)
Tertiary	212 (35.0)
Dead	193 (33.2)

Figure 1:

Estimated travel times to Obafemi Awolowo University Teaching Hospital with geographic distribution of the study cohort. Inset map represents the immediate catchment area of Obafemi Awolowo University Teaching Hospital including the states of Osun, Ekiti and Ondo.

The age and SES of patients presenting with early versus advanced disease were comparable. Level of education was significantly associated with stage at presentation, with a larger proportion of patients with at least tertiary education presenting with early stage disease (51.0%) compared to patients with primary education only (21.9%, p=0.002). Expectedly, there were more deaths during the study period in patients presenting with locally advanced or metastatic disease (38.6% vs 8.6%). The median estimated travel times for patients presenting with locally advanced or metastatic disease were also significantly longer (18.2 vs 46.5 minutes, p=0.010). Unadjusted comparisons for patients presenting with early versus advanced disease are presented in Table 2.

Table 2:

Unadjusted comparisons for patients presenting with early (stage 1–2) versus advanced (stage 3–4) disease

Variable	Stage 1–2 Presentation No. (%) or Median (IQR)	Stage 3–4 Presentation No. (%) or Median (IQR)	p-value
Total	96 (16.4)	490 (83.6)
Age (years)	52 (42–62)	48 (40–58)	0.063
Gender
Female	95 (99.0)	480 (97.9)	0.510
Low SES	71 (74.0)	332 (67.8)	0.297
Education			0.002
None	5 (5.2)	49 (10.1)
Primary	21 (21.9)	168 (34.6)
Secondary	21 (21.9)	114 (23.5)
Tertiary	49 (51.0)	155 (31.9)
Travel Time			0.002
1st Quintile	31 (32.3)	88 (18.0)
2nd Quintile	26 (27.1)	94 (19.2)
3rd Quintile	14 (14.6)	98 (20.0)
4th Quintile	14 (14.6)	113 (23.1)
5th Quintile	11 (11.5)	97 (19.8)
Travel Time (mins)	18.2 (3.6–63.0)	46.5 (8.9–86.2)	0.010
Received Chemotherapy	63 (65.2)	353 (72.0)	0.131
Overall Mortality	8 (8.6)	173 (38.6)	<0.001

After adjusting for age, SES and education, longer estimated travel times to OAUTH remained independently associated with advanced stage at presentation, with the patients in the highest travel time quintile having a 2.8-fold increased odds of an advanced stage presentation relative to the patients in the lowest travel time quintile (95% CI: 1.30–6.11, p=0.006) (Table 3). This association remained when patients with stage 4 disease at presentation were excluded (OR 2.54, p=0.035). The relationship between access and survival was examined with a survival analysis. After adjusting for SES, age, and education, a travel time of ≥ 30 minutes was associated with a significantly increased hazard of death (HR 1.65, 95% CI: 1.17–2.33, p=0.004, Table 4). This is presented in Figure 2. When stage at presentation was included in the adjusted survival analysis the relationship between access and survival remained significant (HR 1.47, 95% CI: 1.04–2.10, p=0.029). A sensitivity analysis was performed by excluding the patients from outside the expected geographic catchment area of OAUTH. In this analysis, the relationship between travel time and stage at presentation was preserved. Patients in the highest travel time quintile had a 2.7-fold increased odds of an advanced stage presentation (p=0.003), and a 1.7-fold increased hazard of death (p=0.003).

Table 3:

Adjusted analyses of the odds of locally advanced or metastatic disease at the time of presentation with increasing travel time

Variable	Adjusted OR (95% CI)	P-value
Travel Time		0.006
1st Quintile	1
2nd Quintile	1.09 (0.58–2.05)
3rd Quintile	2.25 (1.09–4.63)
4th Quintile	2.53 (1.24–5.17)
5th Quintile	2.82 (1.30–6.11)
Low SES	0.49 (0.28–0.85)	0.005
Age	0.97 (0.96–0.99)	0.066
Education		<0.001
None	1
Primary	0.62 (0.22–1.80)
Secondary	0.37 (0.12–1.10)
Tertiary	0.17 (0.06–0.49)

Table 4:

Adjusted analyses of the odds of death with increasing travel time

Variable	Adjusted HR (95% CI)	p-value
Travel time >30 minutes	1.65 (1.17–2.33)	0.004
Low SES	1.19 (0.83–1.72)	0.345
Age	1.00 (0.98–1.01)	0.504
Education		0.365
None	1
Primary	1.25 (0.67–2.35)
Secondary	1.19 (0.60–2.37)
Tertiary	0.88 (0.46–1.72)

Figure 2:

Relationship between travel time to Obafemi Awolowo University Teaching Hospital and survival for patients with breast cancer in South West Nigeria.

Discussion

In this retrospective cohort study, patients with reduced geospatial access to a tertiary referral center (i.e. OAUTH) in South West Nigeria were more than twice as likely to present with locally advanced or metastatic disease. This finding was independent of other risk factors associated with delayed presentation, such as lower SES and lower levels of education. Reduced geospatial access to cancer care was also associated with a significant difference in overall survival. The impact of access on survival persisted even after adjusting for SES, age, education and stage at presentation. This may reflect geospatial variation in receipt of adjuvant care that is not fully captured in our analysis.

At a population level, prior work has demonstrated heterogenous geospatial access to cancer care in Nigeria. However, the impact of access on cancer-related outcomes has not previously be quantified²⁰. This is the first study to utilize a prospectively collected cancer database and GIS modelling to study the relationship between geospatial access to care and breast cancer-specific outcomes in sub-Saharan Africa. The young age of this cohort, as well as the disproportionate burden of advanced disease at the time of presentation is congruent with prior epidemiologic studies^2,5. A multicenter Nigerian study was previously unable to identify an independent association between access to care and stage of presentation, but this work utilized urban versus rural residence as a proxy for geospatial access¹⁰. As this variable does not reflect the spatial relationship between the location of care and the patient’s residence, it is expected to be an imprecise surrogate for spatial access. Additional evidence from South Africa utilizing straight line distance has demonstrated that the risk of advanced breast cancer presentation increased by 0.8% per km from the treatment facility¹⁷. Indeed, a large body of literature supports the relationship between distance and healthcare utilization for inpatient and outpatient care, regardless of age, sex, or illness acuity across sub-Saharan Africa^16,24.

In this study, we utilized the existing road infrastructure to create a time-distance model as an estimate of geospatial access. This is particularly important in settings with poor road infrastructure and security concerns, where distance alone may underestimate the travel burden. As the majority of patients present with advanced disease, understanding the barriers to accessing cancer services is imperative to redressing the disproportionate breast cancer mortality in Nigeria⁸. Although many factors influence an individual’s decision to access healthcare, the geospatial distribution of resources is particularly important in low-and middle-income countries where the costs associated with travel can add significantly to the out-of-pocket expense of healthcare^25,26.

There are several limitations of this study. Owing to the small number of patients with known receptor status, we were unable to include that variable in the analysis. However, we adjusted for variables known to be associated with advanced stage presentation, including, SES, age and education^10,16. The relationship between geospatial access and income is more complex than adjusted for by our simple economic analysis. Individuals with sufficient means will be less constrained by physical distance. To evaluate this potentially confounding variable, we conducted a sensitivity analysis by excluded patients not residing in the catchment area of the hospital. These patients would have bypassed another tertiary care facility to access care at OAUTH. In this sub-analysis, the relationship between travel time and stage at presentation was preserved. Our relatively short median follow-up highlights the challenges of cancer surveillance and prospective database management in lower-resource settings. A large proportion of patients are unable to return to their primary treatment facility for routine follow-up and any additional data collection is resource intensive. This may underestimate mortality for those with a significant distance to travel who are more likely to be lost to follow-up. Tertiary care facilities may also see a higher proportion of late stage disease as patients with more advanced presentation are referred for specialist consultation from smaller secondary facilities. We were unable to account for the volume of breast cancer care that is provided at other facilities within the catchment area or for delay in presentation due to secondary referral. However, the paucity of surgical capacity in rural Nigeria and the similar stage distribution of our cohort to other published series, suggests that our cohort is likely an accurate representation of the pattern of access for breast cancer care in this region of Southwest Nigeria^5,27.

Conclusion

As the burden of breast cancer increases in Nigeria, improving access to high-quality cancer care is a priority. The Federal Ministry of Health endorses a cancer control plan that identifies breast cancer as a priority for investment, including the creation of a network of comprehensive cancer centers²⁸. However, simply bolstering the capacity of tertiary care facilities and comprehensive cancer centers may contribute to divergent outcomes based on place of residence and income. An emphasis on expanding universal health insurance and integrating primary and secondary health care facilities in cancer delivery will be important. As the country’s breast cancer control system matures, measuring access will be essential to ensure equitable health policy and development.