Abstract
Background:
Many rural communities have lost local access to procedural care, and although rural surgical services have endured in some regions, questions regarding quality and safety of care have persisted. Using retrospective observational data, we sought to compare adverse outcomes of the most common surgical procedures performed at rural facilities in British Columbia and outcomes by provider specialty. Our objective was to show whether the efficacy of surgical care at rural facilities is comparable to that of larger referral facilities and whether family physicians with enhanced surgical skills (FPESS) have outcomes comparable to those of specialists at referral facilities for low-morbidity patients.
Methods:
We included patients who had a colonoscopy, hernia repair, appendectomy, or cesarean delivery at 1 of 7 rural hospitals in BC that participated in the Rural Surgical and Obstetrics Networks of BC and corresponding referral facilities between 2016 and 2021. To control for differences in the acuity of patients between facility types, we risk stratified data by patient comorbidity level, in addition to adjusting for other demographic differences using multivariable Firth logistic regression analysis. We also compared the outcomes of FPESS with those of regional specialists for low-acuity patients in a similar manner. We calculated adjusted odds ratios (ORs), used tests of noninferiority to obtain p values for the adjusted ORs, and calculated E-values to estimate the extent to which our findings could be due to other unmeasured confounding.
Results:
Most surgical procedures at rural hospitals were performed by FPESS (n = 4403, 34.9%) and visiting general surgeons (n = 7317, 57.9%). We found that the quality of care at rural facilities was at least equivalent to the quality at referral facilities in rural BC for colonoscopy, hernia repair, and appendectomy, and that FPESS had outcomes at least equivalent to those of specialists for low-acuity patients.
Conclusion:
Our findings provide evidence in favour of the efficacy of rural procedural care at BC facilities, and although these results are not inherently generalizable to other populations, we believe they illustrate the potential for high-quality rural care for low-acuity procedures in similar settings. These findings are an important step toward documenting rural-specific outcomes and creating attendant benchmarks for rural practice.
Abstract
Contexte:
Plusieurs communautés rurales ont perdu l’accès local aux interventions, et même si des services chirurgicaux ont été maintenus dans certaines régions, des questions persistent quant à la qualité et la sécurité des soins. À partir de données observationnelles rétrospectives, nous avons comparé les complications des interventions les plus courantes effectuées dans les établissements ruraux de la Colombie-Britannique, et les résultats selon la spécialité chirurgicale. Notre objectif était de vérifier si l’efficacité des soins chirurgicaux dans les établissements ruraux se compare à celle des grands centres de référence et si les médecins de famille détenant des compétences avancées en chirurgie (MFCAC) obtiennent des résultats comparables à ceux des spécialistes des centres de référence pour les cas de faible morbidité.
Méthodes:
Nous avons inclus des coloscopies, des cure de hernie, des appendicectomies et des césariennes effectuées dans 1 des 7 hôpitaux ruraux de la Colombie-Britannique membres des réseaux ruraux de chirurgie et d’obstétrique de la province et dans leurs centres de référence respectifs entre 2016 et 2021. Pour tenir compte des différences de gravité des cas entre les types d’établissements, nous avons stratifié le risque en fonction du niveau de comorbidité et d’autres différences démographiques au moyen d’une analyse de régression logistique de Firth multivariée. Nous avons aussi comparé de la même manière les résultats des MFCAC à ceux des spécialistes régionaux pour les cas de faible morbidité. Nous avons calculé les rapports des cotes (RC) ajustés, utilisé des tests de non-infériorité pour obtenir les valeurs p des RC ajustés, et nous présentons les valeurs E pour estimer dans quelle mesure nos observations pourraient être dues à d’autres facteurs de confusion non mesurés.
Résultats:
La plupart des interventions chirurgicales effectuées dans les hôpitaux ruraux l’ont été par des MFCAC (n = 4 403, 34,9 %) et par des chirurgiennes et chirurgiens généraux de passage (n = 7 317, 57,9 %). Nous avons constaté que la qualité des soins dans les hôpitaux ruraux était au moins équivalente à celle des centres de référence en milieu rural en Colombie-Britannique pour la coloscopie, la cure de hernie et l’appendicectomie, et que les résultats obtenus par les MFCAC étaient au moins équivalents à ceux des spécialistes pour les cas de faible morbidité.
Conclusion:
Nos observations confirment l’efficacité des soins interventionnels dans les hôpitaux ruraux de la Colombie-Britannique, et même si ces résultats ne sont pas en soi généralisables à d’autres populations, nous croyons qu’ils témoignent du potentiel de qualité des soins ailleurs en milieu rural pour les cas de faible morbidité. Ces observations représentent une étape importante dans le suivi des résultats spécifiques au milieu rural et l’établissement de critères de performance adaptés.
Historically, low-acuity surgical services were offered in small communities across rural Canada, supported by family physicians in most locations.1 Over the past several decades, as a consequence of the confluence of regionalization, fiscal constraints in sustaining community hospital operating rooms (ORs) and lack of availability of family physicians with enhanced surgical skills (FPESS), many rural communities lost local access to procedural care.2 This has most profoundly affected access to maternity care owing to the loss of cesarean delivery capacity. Loss of local care has also had an impact on diagnostics and screening (e.g., colonoscopy) and created the need for residents requiring other procedures such as hernia repair or appendectomy to travel, sometimes great distances, for care.3 Western and northern Canada have been exceptions to this trend through sustained support for FPESS and family physician anesthetists. Despite the endurance of many rural surgical services, questions regarding quality and safety of care have persisted.
The Rural Surgical and Obstetrics Networks (RSON) initiative was funded to stabilize and sustain low-volume rural surgical sites in British Columbia, with the ultimate goal of supporting local births through immediate backup to cesarean delivery. The distributed, site-level funding envelope focused on supporting clinical coaching, community-determined continuous quality-improvement projects, and remote-presence technology to facilitate provider linkages across geography and facilities and to increase scope and volume.4
In this study, we sought to compare the health outcomes of all patients who received care at rural hospitals that participated in the RSON initiative with the outcomes of patients at regional referral sites. We analyzed rural surgical and obstetric outcomes across 7 sites over 5 years, which provided substantial volume for analysis. Additionally, we compared the outcomes of low-acuity procedures done by FPESS and specialists in regional referral centres. This is a tentative first step to create a robust methodology for considering patient safety and quality, as part of a larger rural surgical-quality framework,4 an essential step in supporting and sustaining rural procedural care.
Methods
Setting and context
We conducted a retrospective cohort study. Health service delivery in BC is organized through 5 geographic health authorities, 1 health authority with oversight of provincial services such as emergency transport and cancer care, and 1 provincial authority with oversight over First Nations health.5 Health authorities provide tiers of service that correspond to population needs based on population density and distance to next service, with small communities supporting community hospitals served by family physicians (e.g., for emergency department coverage and in-patient care) and larger communities supporting procedural care performed by a combination of FPESS supported by local family physician anesthetists and visiting specialists.
Data sources and study cohort
Population-level data used in this analysis included all hospital visits for patients who were admitted to 1 of 7 RSON hospitals and the corresponding referral facility between Jan. 1, 2016, and Mar. 31, 2021. These data were provided by Population Data BC and included the Canadian Institute for Health Information (CIHI) Discharge Abstract Database,6 BC Perinatal Data Registry,7 and Consolidation File (MSP Registration & Premium Billing),8 as well as demographic, geospatial data, and Case Mix Group data contained within the Discharge Abstract Database. We used the Centre for Rural Health Research hospital catchment designation for each patient, which refers to the 1-hour catchments around each hospital, based on the distance of the centroid of the postal code of each patient and the distance in surface travel time to the hospital. While the comparison of surgical outcomes presented in this paper is based on the hospital where the surgery occurred, assigning patients to catchments around each hospital enables calculation of the proportion of patients served locally.9 All data joins and sample sizes for each analysis are described in Figure 1.
Fig. 1.
Flow chart showing missing values and data joins. CCI = Canadian Classification of Health Interventions; CMG = Case Mix Group; DAD = Discharge Abstract Database; ICD-10 = International Statistical Classification of Diseases and Related Health Problems, 10th Revision; PSBC = Perinatal Services BC; RSON = Rural Surgical and Obstetrics Networks. See Related Content tab for accessible version.
We compared rates of surgical adverse events of patients at hospitals participating in RSON with the rates of patients who had the same procedures at a referral hospital, and we compared rates of surgical adverse events of rural FPESS and specialists in regional referral centres for patients with comparable acuity. The primary outcomes were general, anesthesia, and procedure-specific adverse events associated with colonoscopy, appendectomy, hernia repair, and cesarean delivery.
Through mapping of Canadian Classification of Health Interventions (CCI) codes, we identified patients who had any of the 4 low-acuity surgical interventions of interest commonly performed in RSON facilities (Appendix 1, Supplemental Table 2, available at www.canjsurg.ca/lookup/doi/10.1503/cjs.003423/tab-related-content). Using a mapping of International Statistical Classification of Diseases and Related Health Problems, 10th Revision, Canada (ICD-10-CA) diagnosis codes, we identified surgical adverse events (general, anesthesia, and procedure-specific adverse events) associated with each intervention. In addition, we identified provider specialty, patient comorbidity level, procedure type (e.g., laparoscopic v. open), and other factors. For the comparative colonoscopy analysis, we excluded 18 cases performed by general practitioners at referral sites as we were interested only in evaluating the outcomes of FPESS. For cesarean deliveries, we compared FPESS with obstetricians rather than general surgeons, as general surgeons primarily dealt with emergency cesarean deliveries.
Our approach for identifying intervention and diagnosis codes was informed by iterative consultations with rural physicians and review of the literature.10 Some of the adverse outcomes that were identified via clinician consultation (e.g., serious events from the Canadian Hospital Harms framework11) were not included because none occurred at RSON hospitals.
Controlling for confounders
Patients with more severe conditions are more likely to receive care at larger hospitals.12 Patients who travel to referral centres have different characteristics from those who have procedures closer to home; for example, in 1 study, younger patients and male patients were more likely to travel outside of the community for care.13 One way to address this potential bias involves controlling for patient characteristics and procedural factors. We adjusted all models for the following variables: patient age (years), sex (male or female), and socioeconomic decile (range 1–10), as well as procedure-specific confounders (e.g., laparoscopic v. open surgery).
Risk stratification by CIHI comorbidity level
Our analyses are risk-stratified by CIHI comorbidity level to account for additional bias concerning the health status of individual patients that may affect their decision in selecting a referral facility over a rural facility for their care. Differences in the acuity of cases between rural and referral facilities that are not explained by patient age, sex, or income decile may be accounted for using CIHI Case Mix Group Comorbidity Levels (CL).6,14 Comorbidity level ranges from 0 to 4 and is based on the expected cost impact of certain comorbidities on patient stay (length of stay, treatment, or both). Comorbidity refers to a diagnosis identified by ICD-10-CA diagnosis code, other than the one that is most responsible for a patient’s hospital admission. Only patients with other diagnosed conditions that significantly increase resource requirements will have a comorbidity level greater than 0. The higher the comorbidity level, the greater the risk for adverse events. To move to a comorbidity level of 1 or higher, the expected increase in costs as a result of 1 or more comorbidities must be at least 25% higher than the costs for a similar patient without comorbidities. We present results for 4 risk strata, including patients with a comorbidity level of 0; those with a comorbidity level of 0 and 1; those with a comorbidity level of 0, 1, and 2; and those presenting all comorbidity levels.
Data analysis
Facility level
We present proportions of individual adverse events for each index procedure for comparison between rural and referral facilities. We consolidated all adverse events per patient to obtain the incidence of any adverse event, which we used to determine the overall adverse-event proportions for each procedure and facility type. We risk stratified data into several risk groups based on CIHI Case Mix Group CL, which range from 0 to 4.
We fit logistic regression analysis using Firth-penalized15 logistic regression models for each risk strata, with a binary response variable indicating the absence (0) or presence (1) of any adverse outcome for that procedure. Even at the relatively large samples available when using population-level data, rare adverse events can result in inflated estimates due to separation or quasi-separation.16 We therefore chose to use Firth logistic regression15 as it has been shown to have reduced bias and superior power for data with rare events; this is examined in detail using a power analysis provided in Appendix 1, Supplemental Table 3. Explanatory variables in each model include the facility type (RSON v. referral) as well as additional confounding explanatory variables, including age, sex, and income decile of patients, and an explanatory variable indicating laparoscopy for hernia repair and appendectomy data. Patient gravidity and an indicator for emergency versus elective procedures were also included as confounders for analysis of cesarean delivery data. Adjusted odds ratios (ORs) and their corresponding confidence intervals (CIs) were obtained for facility type and are presented alongside noninferiority p values, as well as E-values.
To examine whether the rates of surgical adverse events at RSON facilities are as good as that of referral facilities and whether FPESS had outcomes as good as specialist providers, we used tests of noninferiority.17 We chose the noninferiority margins based on half the 95% CI for the raw difference in proportion of adverse events at referral versus rural facilities.18 Our choice of noninferiority margin is more conservative than a common approach used in clinical trials.19 Detailed descriptions of how we implemented these tests of noninferiority including the choice of an appropriate noninferiority margin are described elsewhere.10
We analyzed each procedure separately as procedure-specific adverse events and confounders were included in the models. Colonoscopy, appendectomy, and hernia repair data spanned the period from Jan. 1, 2016, until Mar. 31, 2021; cesarean birth data from Jan. 1, 2016, until Mar. 31, 2021, were included. As rates of missing values were very low (Figure 1 and Appendix 1, Supplemental Table 1), hospital visits with missing data for age and sex were excluded from logistic regression analyses.
In addition, we present E-values20 calculated from the adjusted ORs and their associated CI limits from these models, allowing us to examine the potential effect of unmeasured confounders. These E-values are interpreted as the minimum strength of association between unmeasured confounders with both the treatment (facility type) and the outcome (adverse events rate) that is required to explain away the effect of the facility type on the rate of adverse events. A higher E-value indicates that a stronger confounding effect will be required to explain away the effect of the treatment variable.
Provider comparison
Using logistic regression for each procedure, we fit a model with a binary response variable indicating the absence (0) or presence (1) of an adverse outcome for that procedure. We included age, sex, and socioeconomic decile as explanatory variables to control for differences in the acuity of cases between FPESS and specialist surgeons. Also included in each are additional explanatory variables for each procedure type (e.g., open v. laparoscopic) and controlled for patient comorbidity.
Patient comorbidity differs between provider specialty and especially between rural and referral site patients. Higher-risk patients may choose to receive care at referral facilities, where they are most likely treated by a specialist rather than a family doctor. To control for differences in the acuity of patients in our comparison of provider specialty, we additionally provide the results of an analysis for a low-risk cohort, as defined by CIHI Case Mix Group CL.14 The CL range from 0 to 4 based on anticipated cost impact incurred (for treatment, length of stay, or both). Comorbidity is defined as any patient condition other than the one most responsible for the hospital admission episode. Although most patients will have a CL of 0, those with diagnosed conditions are likely to increase resource usage more than 25% than the costs for patients without comorbidities. Although we present the results of all risk strata, we have prioritized analysis for patients of comorbidity level 0 to account for the potential of referral bias to larger centres.
Results
During the study period, 10 559 colonoscopies, 999 hernia repairs, 937 cesarean deliveries, and 136 appendectomies were performed at RSON hospitals (Table 1). Most of these procedures were performed by FPESS (n = 4403, 34.9%) and visiting general surgeons (n = 7317, 57.9%). The remaining procedures were done by either local or visiting specialists beyond general surgeons (e.g., gastroenterologists and orthopedic surgeons). Two sites had resident specialists. Volumes of all index procedures, except cesarean delivery, increased over the study period at RSON hospitals, likely as a result of the additional funding and resources available. For example, in January 2021 there were 137 more colonoscopies performed than in January 2016. The COVID-19 pandemic affected procedural volumes owing to required temporary closures (Figure 2). The proportion of patients from RSON communities who were able to have their procedures done locally increased between 2016 and 2021, with absolute differences between 5% (hernia repair) and 14% (appendectomy) (Figure 3).
Table 1.
Patient characteristics by facility type, for each procedure
| Variable | No. (%)* | ||
|---|---|---|---|
| All | Rural | Referral | |
| Hernia repair | n = 7792 | n = 999 | n = 6793 |
| Patient transfers | –† | < 5 (0.2) | 69 (1.0) |
| Local procedures | 5201 (66.7) | 814 (81.5) | 4387 (64.6) |
| Sex, female | 1966 (25.2) | 192 (19.2) | 1774 (26.1) |
| Income decile, mean ± SD | 6.8 ± 9.1 | 6.9 ± 6.3 | 6.8 ± 9.4 |
| Age, yr, mean ± SD | 58.0 ± 16.5 | 55.0 ± 16.3 | 58.4 ± 16.5 |
| Laparoscopic | 832 (10.7) | 74 (7.4) | 758 (11.2) |
| Mean comorbidity level, mean ± SD | 0.08 ± 0.00006 | 0.006 ± 0.00009 | 0.09 ± 0.00007 |
| Appendectomy | n = 2613 | n = 136 | n = 2477 |
| Patient transfers | 46 (1.8) | 5 (3.7) | 41 (1.7) |
| Local procedures | 1692 (64.8) | 117 (86.0) | 1575 (63.6) |
| Sex, female | 1343 (51.4) | 68 (50.0) | 1275 (51.5) |
| Income decile, mean ± SD | 6.8 ± 8.9 | 7.2 ± 4.0 | 6.8 ± 9.1 |
| Age, yr, mean ± SD | 37.8 ± 20.2 | 31.1 ± 18.4 | 38.2 ± 20.3 |
| Laparoscopic | 2016 (77.2) | 65 (47.8) | 1951 (78.8) |
| Mean comorbidity level, mean ± SD | 0.13 ± 0.0002 | 0.03 ± 0.002 | 0.13 ± 0.0002 |
| Colonoscopy | n = 51 968 | n = 10 559 | n = 41 409 |
| Patient transfers | 175 (0.3) | 20 (0.2) | 155 (0.4) |
| Local procedures | 39 623 (76.2) | 8427 (79.8) | 31 196 (75.3) |
| Sex, female | 24 365 (46.9) | 4937 (46.8) | 19 428 (46.9) |
| Income decile, mean ± SD | 6.9 ± 9.2 | 6.7 (8.1) | 7.0 (9.4) |
| Age, yr, mean ± SD | 63.4 ± 11.2 | 62.2 (11.3) | 63.7 (11.2) |
| Mean comorbidity level, mean ± SD | 0.02 ± 0.000005 | 0.008 ± 0.00001 | 0.03 ± 0.000007 |
| Cesarean delivery | n = 5972 | n = 937 | n = 5035 |
| Patient transfers | 120 (2.0) | 25 (2.7) | 95 (1.9) |
| Local procedures | 4454 (74.6) | 813 (86.8) | 3641 (72.3) |
| Income decile, mean ± SD | 6.7 ± 8.1 | 7.1 ± 5.6 | 6.6 ± 8.5 |
| Age, yr, mean ± SD | 30.9 ± 5.3 | 31.0 ± 5.4 | 30.9 ± 5.3 |
| Mean comorbidity level, mean ± SD | 0.17 ± 0.00008 | 0.10 ± 0.0004 | 0.18 ± 0.0001 |
| Emergency cesarean deliveries | 3711 (62.1) | 577 (61.6) | 3134 (62.2) |
| Gravidity, mean ± SD | 2.5 ± 0.0003 | 2.3 ± 0.002 | 2.4 ± 0.0004 |
SD = standard deviation.
Unless stated otherwise.
Value is omitted as it would expose a count of less than 5, against data privacy protocol.
Fig. 2.
Quarterly colonoscopy volumes at Rural Surgical and Obstetrics Network (RSON) hospitals.
Fig. 3.
Proportion of local procedures, over time.
Patient characteristics by facility type and provider specialty
Patient characteristics by facility type and provider specialty are shown in Table 2. Patients who had hernia repairs at RSON, compared with those at referral hospitals, were, on average, younger and less likely to be female, and had lower comorbidity levels. This was consistent for those who received care from FPESS compared with specialists as well. Patients who had appendectomies at RSON hospitals had lower comorbidity levels, were younger, and were much less likely to have laparoscopic surgery than patients at referral centres. Again, this was consistent for the FPESS cohort in the provider-specialty analysis. Patients who had colonoscopies at rural hospitals, compared with referral centres, had similar characteristics (similar incomes, ages, and sex distribution) but had lower comorbidity levels, again reflected in the FPESS cohort in the provider outcomes. Child-bearing people who delivered by cesarean at RSON hospitals had very similar patient characteristics as those at referral facilities. Both referral centre and RSON cohorts had a rate of emergency cesarean delivery of around 62% (Table 1). However, although the FPESS cohort had a lower rate of adverse events than the specialists (obstetricians and general surgeons), they also had patients with a lower mean comorbidity level.
Table 2.
Patient characteristics by provider type at rural facilities and referral facilities, by provider specialty
| Variable | Total | FPESS | General surgeon or obstetrician | FPESS low risk strata | General surgeon or obstetrician low risk strata |
|---|---|---|---|---|---|
| Hernia repair | |||||
| Total cases, no. | 7704 | 350 | 7354 | 347 | 7028 |
| Adverse event rate, % | 3.4 | 0.9 | 3.6 | 0.9 | 1.6 |
| Patient transfer, % | 0.9 | 0.0 | 1.0 | 0.0 | 0.5 |
| Laparoscopic, % | 10.7 | 3.1 | 11.0 | 3.2 | 11.1 |
| Age, mean ± SD | 58.2 ± 16.4 | 53.2 ± 16.6 | 58.4 ± 16.3 | 53.1 ± 16.6 | 58.0 ± 16.3 |
| Sex, female, % | 24.8 | 16.6 | 25.2 | 16.4 | 24.2 |
| Socioeconomic decile, mean ± SD | 6.83 ± 9.1 | 6.84 ± 3.5 | 6.83 ± 9.3 | 6.85 ± 3.5 | 6.82 ± 9.1 |
| Mean comorbidity level (0–4) | 0.079 | 0.011 | 0.083 | 0 | 0 |
| Appendectomy | |||||
| Total cases, no. | 2601 | 86 | 2515 | 84 | 2297 |
| Adverse event rate, % | 8.1 | 4.7 | 8.2 | 2.4 | 3.2 |
| Patient transfer, % | 1.8 | 3.5 | 1.7 | 2.4 | 1.2 |
| Laparoscopic, % | 77.2 | 39.5 | 78.5 | 39.3 | 80.4 |
| Age, mean ± SD | 37.8 ± 20.2 | 27.6 ± 17.4 | 38.2 ± 20.2 | 27.5 ± 17.5 | 37.1 ± 19.8 |
| Sex, female, % | 51.2 | 51.2 | 51.2 | 51.2 | 51.1 |
| Socioeconomic decile, mean ± SD | 6.8 ± 8.9 | 7.3 ± 3.8 | 6.8 ± 9.0 | 7.3 ± 3.8 | 6.9 ± 9.1 |
| Mean comorbidity level (0–4) | 0.127 | 0.05 | 0.13 | 0 | 0 |
| Colonoscopy | |||||
| Total cases, no. | 47800 | 3100 | 44700 | 3089 | 44123 |
| Adverse event rate, % | 1.0 | 0.6 | 1.0 | 0.5 | 0.5 |
| Patient transfer, % | 0.3 | 0.2 | 0.3 | 0.1 | 0.2 |
| Sigmoidoscopy, % | 3.5 | 1.2 | 3.7 | 1.2 | 3.5 |
| Age, mean ± SD | 63.6 ± 11.2 | 60.9 ± 11.0 | 63.7 ± 11.0 | 60.9 ± 10.6 | 63.6 ± 11.1 |
| Sex, female, % | 46.8 | 44.0 | 47.0 | 44.1 | 47.0 |
| Socioeconomic decile, mean ± SD | 7.0 ± 9.4 | 7.7 ± 9.7 | 6.9 ± 9.4 | 7.7 ± 9.7 | 6.9 ± 9.3 |
| Mean comorbidity level (0–4) | 0.0239 | 0.008 | 0.025 | 0 | 0 |
| Cesarean delivery | |||||
| Total cases, no. | 5679 | 1110 | 4569 | 1013 | 3906 |
| Adverse event rate, % | 2.1 | 1.4 | 2.2 | 1.2 | 1.6 |
| Patient transfer, % | 1.9 | 2.3 | 1.8 | 2.0 | 1.7 |
| Emergency rate, % | 61.4 | 59.8 | 61.8 | 57.1 | 57.2 |
| Age, mean ± SD | 30.9 ± 5.3 | 30.8 ± 5.4 | 31.0 ± 5.3 | 30.9 ± 5.3 | 31.0 ± 5.2 |
| Socioeconomic decile, mean ± SD | 6.6 ± 7.9 | 7.6 ± 9.7 | 6.3 ± 7.4 | 7.7 ± 9.9 | 6.3 ± 7.0 |
| Mean comorbidity level (0–4) | 0.169 | 0.096 | 0.186 | 0 | 0 |
FPESS = family physicians with enhanced surgical skills; SD = standard deviation.
Adverse events for each procedure
Patients who had hernia repairs experienced lower rates of adverse events at rural hospitals (1.2%) than at referral hospitals (3.8%). Hematoma and hemorrhage were the most commonly experienced hernia repair adverse events at RSON hospitals (0.7%) regardless of comorbidity level (Table 3). The overall rate of appendectomy adverse events was 4.4% at RSON and 8.3% at referral facilities. The most common appendectomy adverse events at RSON hospitals were gastrointestinal adverse events (2.9%) and surgical site infections (2.9%) (Table 4). Adverse events related to colonoscopies were uncommon at both RSON and referral facilities (0.5% and 1.2% respectively) (Table 5). Cesarean delivery adverse events were similar at RSON and referral hospitals (1.8% v. 2.1%) (Table 6). Rates of adverse events for each index procedure were lower at RSON hospitals than at referral sites, with absolute differences ranging from 0.3% for cesarean delivery to 3.9% for appendectomy (Tables 3–6).
Table 3.
Adverse events for hernia repairs
| Hernia repair adverse event | Procedures with adverse events, % | |||
|---|---|---|---|---|
| Rural | Referral | Rural (comorbidity level 0) | Referral (comorbidity level 0) | |
| Any hernia complication | 1.2 | 3.8 | 1.2 | 1.7 |
| Sepsis | 0.0 | 0.3 | 0.0 | 0.0 |
| Myocardial infarction | 0.0 | 0.1 | 0.0 | 0.0 |
| Pulmonary embolism | 0.0 | 0.1 | 0.0 | 0.1 |
| Cerebrovascular accident (stroke) | 0.1 | 0.1 | 0.1 | 0.0 |
| Aspiration pneumonia | 0.0 | 0.2 | 0.0 | 0.0 |
| Problems with airways in postanesthesia recovery, excluding aspiration pneumonia | 0.1 | 0.1 | 0.1 | 0.0 |
| Bowel injury | 0.1 | 0.4 | 0.1 | 0.2 |
| Acute renal failure | 0.0 | 0.7 | 0.0 | 0.1 |
| Hematoma and hemorrhage | 0.7 | 0.9 | 0.7 | 0.5 |
| Cut, puncture, perforation, or hemorrhage | 0.1 | 0.6 | 0.1 | 0.4 |
| Wound disruption | 0.0 | 0.7 | 0.0 | 0.2 |
| Surgical site infection | 0.2 | 1.0 | 0.2 | 0.4 |
| Foreign body | 0.0 | 0.0 | 0.0 | 0.0 |
| Deep vein thrombosis | 0.0 | 0.0 | 0.0 | 0.0 |
| Bladder injury | 0.0 | 0.1 | 0.0 | 0.0 |
| Failed tracheal intubation | 0.0 | 0.0 | 0.0 | 0.0 |
| Cardiac pulmonary resuscitation | 0.0 | 0.1 | 0.0 | 0.0 |
Table 4.
Adverse events for appendectomies
| Appendectomy adverse event | Procedures with adverse events, % | |||
|---|---|---|---|---|
| Rural | Referral | Rural (comorbidity level 0) | Referral (comorbidity level 0) | |
| Any appendectomy complication | 4.4 | 8.3 | 3.0 | 3.2 |
| Sepsis | 0.0 | 0.9 | 0.0 | 0.3 |
| Gastrointestinal complication | 2.9 | 4.0 | 2.2 | 0.8 |
| Acute renal failure | 0.0 | 0.5 | 0.0 | 0.0 |
| Wound disruption | 0.0 | 0.2 | 0.0 | 0.0 |
| Surgical site infection | 2.9 | 3.0 | 2.2 | 1.8 |
| Myocardial infarction | 0.0 | 0.0 | 0.0 | 0.0 |
| Cerebrovascular accident (stroke) | 0.0 | 0.1 | 0.0 | 0.0 |
| Aspiration pneumonia | 0.0 | 0.1 | 0.0 | 0.0 |
| Problems with airways in postanesthesia recovery, excluding aspiration pneumonia | 0.7 | 0.1 | 0.0 | 0.0 |
| Bowel injury | 0.0 | 0.4 | 0.0 | 0.1 |
| Bladder injury | 0.0 | 0.0 | 0.0 | 0.0 |
| Cut, puncture, perforation, or hemorrhage | 0.0 | 0.8 | 0.0 | 0.6 |
| Pulmonary embolism | 0.0 | 0.1 | 0.0 | 0.0 |
| Deep vein thrombosis | 0.0 | 0.0 | 0.0 | 0.0 |
| Abscess drainage | 0.0 | 0.1 | 0.0 | 0.1 |
| Cardiac pulmonary resuscitation | 0.0 | 0.0 | 0.0 | 0.0 |
Table 5.
Adverse events for colonoscopies
| Colonoscopy adverse event | Procedures with adverse events, % | |||
|---|---|---|---|---|
| Rural | Referral | Rural (comorbidity level 0) | Referral (comorbidity level 0) | |
| Any colonoscopy complication | 0.5 | 1.2 | 0.4 | 0.6 |
| Sepsis | 0.0 | 0.1 | 0.0 | 0.0 |
| Myocardial infarction | 0.0 | 0.1 | 0.0 | 0.0 |
| Pulmonary embolism | 0.1 | 0.1 | 0.0 | 0.0 |
| Cerebrovascular accident (stroke) | 0.0 | 0.1 | 0.0 | 0.0 |
| Deep vein thrombosis | 0.0 | 0.03 | 0.0 | 0.0 |
| Aspiration pneumonia | 0.0 | 0.1 | 0.0 | 0.0 |
| Bowel injury | 0.1 | 0.1 | 0.1 | 0.0 |
| Acute renal failure | 0.1 | 0.4 | 0.1 | 0.1 |
| Heavy bleeding postprocedure | 0.2 | 0.2 | 0.2 | 0.2 |
| Cut, puncture, perforation, or hemorrhage | 0.1 | 0.1 | 0.1 | 0.1 |
| Wound disruption | 0.0 | 0.0 | 0.0 | 0.0 |
| Surgical site infection | 0.0 | 0.1 | 0.0 | 0.0 |
| Foreign body | 0.0 | 0.0 | 0.0 | 0.0 |
| Problems with airways in postanesthesia recovery, excluding aspiration pneumonia | 0.0 | 0.0 | 0.0 | 0.0 |
| Bladder injury | 0.0 | 0.0 | 0.0 | 0.0 |
| Blood transfusion | 0.0 | 0.0 | 0.0 | 0.0 |
| Cardiac pulmonary resuscitation | 0.0 | 0.0 | 0.0 | 0.0 |
Table 6.
Adverse events for cesarean deliveries
| Cesarean delivery adverse event | Procedures with adverse events, % | |||
|---|---|---|---|---|
| Rural | Referral | Rural (comorbidity level 0) | Referral (comorbidity level 0) | |
| Any cesarean delivery complication | 1.8 | 2.1 | 1.2 | 1.6 |
| Bladder, urethra injury | 0.5 | 0.5 | 0.4 | 0.42 |
| Bladder injury | 0.0 | 0.1 | 0.0 | 0.1 |
| Cut, puncture, perforation, or hemorrhage | 0.1 | 0.6 | 0.1 | 0.5 |
| Acute renal failure | 0.3 | 0.2 | 0.2 | 0.1 |
| Pulmonary embolism | 0.1 | 0.1 | 0.0 | 0.1 |
| Cerebrovascular accident (stroke) | 0.0 | 0.0 | 0.0 | 0.0 |
| Surgical site infection | 0.5 | 0.8 | 0.4 | 0.4 |
| Puerperal sepsis, including endometritis | 0.3 | 0.4 | 0.2 | 0.2 |
| Aspiration pneumonia | 0.0 | 0.0 | 0.0 | 0.0 |
| Sepsis | 0.1 | 0.0 | 0.0 | 0.0 |
| Deep vein thrombosis | 0.0 | 0.0 | 0.0 | 0.0 |
| Foreign body | 0.0 | 0.0 | 0.0 | 0.0 |
| Wound disruption | 0.2 | 0.1 | 0.1 | 0.0 |
| Blood transfusion | 0.2 | 0.0 | 0.2 | 0.0 |
| Intraoperative hysterectomy | 0.0 | 0.1 | 0.0 | 0.0 |
| Cardiac pulmonary resuscitation | 0.0 | 0.0 | 0.0 | 0.0 |
Analysis of comparative outcomes between FPESS and specialists (general surgeons and obstetricians) for all procedures showed that after controlling for age, sex, and income decile, there are lower odds of surgical adverse events for FPESS. Tests of noninferiority exhibit significant (p < 0.05) or borderline significant (p < 0.07) p values for all procedures, meaning that it is unlikely that FPESS will exhibit adverse events rates that are worse than those of a specialist for the index procedures. These test results depend largely on our choice of noninferiority margin (Table 7), and we consider our choice to be relatively conservative based on existing literature.
Table 7.
Logistic regression analysis results across risk strata for comparisons between rural general practitioners and specialists (obstetricians for cesarean delivery, general surgeons for all other procedures)
| Comorbidity level; procedure | General practitioner v. specialist adjusted OR* (95% CI) | p value | NI margin, % (OR) | NI p value |
|---|---|---|---|---|
| Patients with comorbidity level of 0 | ||||
| Hernia repair | 0.78 (0.24–2.48) | 0.5 | 0.9 (1.56) | 0.1 |
| Appendectomy | 0.68 (0.16–2.89) | 0.4 | 2.1 (1.69) | 0.1 |
| Colonoscopy | 1.19 (0.71–1.99) | 0.6 | 0.1 (1.28) | 0.4 |
| Cesarean delivery | 0.77 (0.41–1.44) | 0.4 | 0.6 (1.38) | 0.04 |
| Patients with comorbidity level of 0 or 1 | ||||
| Hernia repair | 0.57 (0.18–1.81) | 0.4 | 1.2 (1.56) | 0.04 |
| Appendectomy | 0.46 (0.14–1.52) | 0.4 | 3.4 (1.60) | 0.02 |
| Colonoscopy | 1.01 (0.62–1.66) | 0.5 | 0.2 (1.29) | 0.2 |
| Cesarean delivery | 0.69 (0.40–1.19) | 0.4 | 0.6 (1.32) | 0.01 |
| All patients (comorbidity levels 0, 1, 2, 3, and 4) | ||||
| Hernia repair | 0.41 (0.13–1.30) | 0.4 | 1.9 (1.56) | 0.01 |
| Appendectomy | 0.45 (0.16–1.26) | 0.4 | 4.0 (1.56) | 0.01 |
| Colonoscopy | 0.72 (0.45–1.16) | 0.4 | 0.4 (1.36) | 0.004 |
| Cesarean delivery | 0.65 (0.38–1.11) | 0.3 | 0.8 (1.37) | 0.003 |
CI = confidence interval; NI = noninferiority; OR = odds ratio.
All models adjusted for age, sex, and socioeconomic decile. Hernia repair model also adjusted for laparoscopic (v. open) and use of tissue (v. none). Appendectomy model also adjusted for laparoscopic (v. open). Colonoscopy model also adjusted for sigmoidoscope (v. colonoscope). Cesarean delivery model also adjusted for emergency surgery and gravidity.
Multivariable results
Colonoscopy
Controlling for patient age, sex, income decile, and colonoscopy type (sigmoidoscope v. colonoscope), we found that, among patients with a comorbidity level of 0, the adjusted odds of colonoscopy-related adverse events at RSON facilities were 0.85 times that of referral facilities; that is, 15% lower (Table 8). Using a test of noninferiority, with a p value of 0.02, and controlling for confounders, we found that the odds of a colonoscopy patient experiencing an adverse event at an RSON facility were at least equivalent to that of referral facilities, meaning that it is unlikely that a patient will more often experience a colonoscopy adverse event at an RSON facility than at a referral facility.
Table 8.
Logistic regression analysis results across risk strata, including noninferiority tests on adjusted odds ratios for Firth-penalized logistic regression
| Comorbidity level; procedure | Rural v. referral adjusted OR* (95% CI) | p value | NI margin, % (OR)† | NI p value‡ | E-value, point estimate | E-value, 95% CI upper limit |
|---|---|---|---|---|---|---|
| Patients with comorbidity level of 0 | ||||||
| Hernia repair | 0.82 (0.43–1.43) | 0.5 | 0.6 (1.37) | 0.07 | 1.74 | 1.00 |
| Appendectomy | 0.71 (0.19–1.87) | 0.5 | 1.2 (1.38) | 0.2 | 2.17 | 1.00 |
| Colonoscopy | 0.85 (0.61–1.16) | 0.3 | 0.1 (1.24) | 0.02 | 1.63 | 1.00 |
| Cesarean delivery | 0.78 (0.38–1.45) | 0.5 | 0.6 (1.39) | 0.07 | 1.88 | 1.00 |
| Patients with comorbidity level of 0 or 1 | ||||||
| Hernia repair | 0.59 (0.31–1.01) | 0.05 | 1.0 (1.42) | < 0.001 | 2.78 | 1.00 |
| Appendectomy | 0.42 (0.14–0.99) | 0.048 | 2.6 (1.46) | 0.003 | 4.19 | 1.11 |
| Colonoscopy | 0.69 (0.51–0.93) | 0.01 | 0.2 (1.29) | < 0.001 | 2.26 | 1.36 |
| Cesarean delivery | 0.85 (0.48–1.42) | 0.5 | 0.6 (1.33) | 0.09 | 1.63 | 1.00 |
| Patients with comorbidity level of 0, 1, or 2 | ||||||
| Hernia repair | 0.47 (0.25–0.81) | 0.005 | 1.3 (1.45) | < 0.001 | 3.68 | 1.78 |
| Appendectomy | 0.33 (0.11–0.79) | 0.009 | 3.3 (1.49) | < 0.001 | 5.51 | 1.85 |
| Colonoscopy | 0.58 (0.42–0.77) | < 0.001 | 0.3 (1.32) | < 0.001 | 2.84 | 1.92 |
| Cesarean delivery | 0.91 (0.53–1.49) | 0.7 | 0.6 (1.30) | 0.2 | 1.43 | 1.00 |
| All patients (comorbidity levels 0,1,2,3, and 4) | ||||||
| Hernia repair | 0.37 (0.20–0.64) | < 0.001 | 1.7 (1.48) | < 0.001 | 4.85 | 2.50 |
| Appendectomy | 0.38 (0.14–0.83) | 0.01 | 3.4 (1.47) | < 0.001 | 4.70 | 1.70 |
| Colonoscopy | 0.46 (0.34–0.60) | < 0.001 | 0.4 (1.37) | < 0.001 | 3.77 | 2.72 |
| Cesarean delivery | 0.88 (0.51–1.42) | 0.6 | 0.6 (1.31) | 0.1 | 1.53 | 1.00 |
CI = confidence interval, NI = noninferiority; OR = odds ratio.
All models adjusted for age, sex, and socioeconomic decile. Hernia repair model also adjusted for laparoscopic (v. open) and use of tissue (v. none). Appendectomy model also adjusted for laparoscopic (v. open). Colonoscopy model also adjusted for sigmoidoscope (v. colonoscope). Cesarean delivery model also adjusted for emergency surgery and gravidity.
Noninferiority margins presented as a proportion as well as an OR limit.
p values for tests of noninferiority are included, as well as noninferiority margins in the form of an acceptable margin above the complication rate for referral facilities.
When considering the adjusted odds of colonoscopy-related adverse events for FPESS compared with specialists, we found that for patients with a comorbidity level of 0, the adjusted odds were 1.19 (95% CI 0.71–1.99); for patients with a comorbidity level of 0 or 1, the adjusted odds were 1.01 (95% CI 0.62–1.66); and for all comorbidity levels (0–4), the adjusted odds were 0.72 (95% CI 0.45–1.16) (Table 7).
There were no significant differences in the noninferiority test for comorbidity level 0, with a p value of 0.4, nor in comorbidity levels 0–1 (p = 0.2). However, there were significant differences in the noninferiority p value for comorbidity levels 0–4 (p = 0.004).
Appendectomy
Controlling for patient age, sex, income decile, and procedure type (laparoscopy v. open), we found that the odds of appendectomy-related adverse events among patients with a comorbidity level of 0 at RSON facilities were 0.71 times that of referral facilities, that is, 29% lower (Table 8). Based on a noninferiority test with a p value of 0.2, we were unable to show evidence for or against whether rates of appendectomy-related adverse events at RSON facilities were at least equivalent to those of referral facilities. Among patients from the second-lowest risk strata (i.e., those with a comorbidity level of 0 or 1), those at RSON facilities had 58% lower odds of experiencing appendectomy-related adverse events, with a noninferiority p value of 0.003, showing strong evidence in favour of a similar odds of appendectomy-related adverse events between rural and referral facilities, for the second-lowest risk strata.
When considering the adjusted odds of appendectomy-related adverse events for FPESS compared with specialists, we found that for patients with a comorbidity level of 0, the adjusted odds were 0.68 (95% CI 0.16–2.89); for patients with a comorbidity level of 0 or 1, the adjusted odds were 0.46 (95% CI 0.14–1.52); and for all comorbidity levels (0–4), the adjusted odds were 0.45 (95% CI 0.16–1.26). There were no significant differences in the noninferiority test for comorbidity level 0, with a p value of 0.1. However, there were significant differences in the noninferiority p value for comorbidity levels 0–1 (0.02) and levels 0–4 (p = 0.009) (Table 7).
Hernia repair
Controlling for patient age, sex, income decile, and hernia-repair type (laparoscopy v. open), we found that the odds of hernia repair–related adverse events at RSON facilities were 0.82 times that of referral facilities; that is, 18% lower for patients with comorbidity level 0 (Table 8). Using a test of noninferiority, with a p value of 0.07, we found no differences in hernia repair–related adverse events at RSON compared with referral facilities. Among patients from the second-lowest risk strata (i.e., those with a comorbidity level of 0 or 1), those at RSON facilities had 41% lower odds of experiencing hernia repair–related adverse events, with a noninferiority p value of less than 0.001, showing strong evidence in favour of a similar odds of hernia repair–related adverse events between rural and referral facilities.
When considering the adjusted odds of hernia repair–related adverse events for FPESS compared with specialists, we found that for patients with a comorbidity level of 0, the adjusted odds were 0.78 (95% CI 0.24–2.48); for patients with a comorbidity level of 0 or 1, the adjusted odds were 0.57 (95% CI 0.18–1.81); and for all comorbidity levels (0–4), the adjusted odds were 0.41 (95% CI 0.13–1.30). There were no significant differences in the noninferiority test for comorbidity level 0, with a p value of 0.1. However, there were significant differences in the noninferiority p value for comorbidity levels 0–1 (0.04) and levels 0–4 (0.01).
Cesarean delivery
Controlling for patient age, income decile, planned versus emergency cesarean delivery, and gravidity, we found that the odds of cesarean delivery–related adverse events at RSON facilities were 0.78 times that of referral facilities; that is, 22% lower (Table 8). Using a test of noninferiority, with a p value of 0.07, we found no evidence for or against whether rates of cesarean delivery–related adverse events at RSON facilities were at least equivalent to that of referral facilities. Higher procedural volume is required to evaluate whether cesarean birth outcomes are similar between rural and referral facilities.
When considering the adjusted odds of cesarean delivery–related adverse events for FPESS compared with specialists, we found that for patients with a comorbidity level of 0, the adjusted odds were 0.77 (95% CI 0.41–1.44); for patients with a comorbidity level of 0 or 1, the adjusted odds were 0.69 (95% CI 0.40–1.19); and for all comorbidity levels (0–4), the adjusted odds were 0.65 (95% CI 0.38–1.11). There were no significant differences in the noninferiority test for comorbidity level 0, with a p value of 0.04. However, there were significant differences in the noninferiority p value for comorbidity levels 0–1 (p = 0.01) and levels 0–4 (p = 0.003) (Table 7).
Logistic regression results for provider comparison
For all procedures, after controlling for age, sex, and income decile, we observed lower odds of surgical adverse events for FPESS than for general surgeons (or obstetricians). Tests of noninferiority exhibited significant (< 0.05) or borderline significant (< 0.07) p values for all procedures, meaning that it is unlikely that FPESS will have rates of adverse events that are worse than those of a specialist for the index procedures. These test results depend largely on our choice of noninferiority margin (Table 7), and we believe our choice to be relatively conservative based on existing literature.
Discussion
Population health outcomes are an essential part of a patient-safety and quality framework in health care and have historically underscored most assessment,21 though more recently alongside process measures such as team function and provider retention.22 The challenges of robust analysis across low procedural volume characteristics of rural health services, however, are considerable and include lack of rural-specific measures and lack of appropriate methodologies to specifically address low case volume.23 Taken together with the onerous clinical responsibilities of rural providers and the lack of local peer support for engaging in quality-improvement projects, there are gaps in reporting rural health outcomes.
Although methodological issues exist, we have attempted to address low procedural volume by pooling data across all RSON sites over 4 years and using composite measures. Most importantly, however, we have endeavoured to create a fair comparison between the outcomes at RSON and referral sites by including analysis of patients with a comorbidity level of 0 and prioritizing a statistical approach that accounts for unrecognized confounders. Likewise, we have applied this approach to the inherently challenging comparison by provider type, owing to the possibility of unrecognized referral bias.
The adjusted ORs from logistic regression analyses in Table 7 do not show any significant difference in the rates of adverse events between general practitioners and specialists. Noninferiority tests for the risk strata including all comorbidity levels exhibited significant noninferiority between general practitioners and specialists. For all procedures except colonoscopy, we found significant noninferiority for patients with comorbidity levels 0 and 1. E-values are included but are not necessary as there were no significant ORs in the logistic regression analyses.
Within the context of the approach presented here, we assert that these findings are an important step toward documenting rural-specific and rural provider–specific outcomes and creating attendant benchmarks for rural practice. Using noninferiority testing and controlling for income, age, sex, and complexity of procedure, and stratifying by patient comorbidity level, we found strong evidence that colonoscopy adverse-event proportions at RSON facilities were noninferior to those at referral facilities among patients with a comorbidity level of 0. Likewise, this was found through the provider comparison analysis. Similarly, we found strong evidence that hernia repair and appendectomy cases had similar proportions of adverse events at rural and referral facilities for patients with a comorbidity level of 0 or 1. We did not find significant p values for cesarean delivery for any of the risk strata, although they were approaching significance; this underscores the need for further analysis with increased volume at rural facilities. Additionally, factors contributing to cesarean surgical risk, including gravidity, age, and proportion of emergency surgeries, were similarly distributed across rural and referral facilities, further highlighting the need for more data.
Previous conclusions from research concerning the efficacy of rural care are mixed, with some authors finding evidence in favour of the quality of cesarean deliveries, colonoscopies, hernia repairs, and appendectomies performed by family physicians or visiting specialists in rural settings,24–29 while others found a lower or mixed quality of care from rural practitioners for colonoscopy, appendectomy, and cesarean delivery.30–34 Likewise, there is mixed evidence regarding the volume-to-outcomes relation for low-acuity procedures, with the exception of cesarean delivery. Drukker and colleagues35 found that patients cared for by providers who performed or supervised more than 48 cesarean deliveries a year (the median volume of the cohort) had fewer urinary and gastrointestinal injuries, less blood loss, and a reduced need for prolonged maternal hospital stay. Overall there were significant risk reductions for cesarean delivery performed or supervised by obstetricians with annual volumes of 21–60 (23%), 61–120 (25%), and more than 120 (34%) compared with those who performed or supervised fewer than 20 per year. However, it should be noted that the study cohort included specialist obstetricians as opposed to FPESS.
Results are more equivocal for other surgeries and procedures. O’Connell and colleagues36 found no significant differences in appendectomy outcomes based on case volume (< 20 cases/yr v. > 62 cases/yr), although they noted the propensity by those in the low-volume cohort to perform laparoscopic as opposed to open procedures. Wei and colleagues37 found a discrepancy in appendectomy outcomes between low-volume surgeons (≤ 65 cases/yr) and those with higher volume (121–190 cases/yr), the former experiencing increased morbidity and perforation compared with the latter. Likewise, several studies noted the need for recurrent hernia repair among lower-volume surgeons,38–46 but the definition of low volume varied considerably across these studies, ranging from less than 5 in one study46 to 100 in another.42 Köckerling and colleagues45 noted that although differences exist based on procedural volume for hernia repair, both lower- and higher-volume groups offered high-quality care. Similarly, evidence on volume to outcomes for colonoscopy is conflicting, with Pace and colleagues47 finding an inverse relation between higher volume of procedures and adverse events, and no differences in adenoma detection rates, whereas Baxter and colleagues48 found a lack of association between volume and postcolonoscopy colorectal cancer detection. In contrast, Singh and colleagues49 found higher rates of adverse events among low-volume providers (< 200/yr) (5.4/1000 v. 2.7/1000).
Our findings contribute to this literature and provide evidence in favour of the high efficacy of rural procedural care at BC facilities participating in the RSON initiative. Although these results are not inherently generalizable to other populations, we believe they illustrate the potential for high-quality rural care for low-acuity procedures, performed both by FPESS and visiting or resident specialists in similar settings.
Limitations
The consolidation of adverse events into overall proportions substantially increased the power of our analyses, but as a result, the regression analyses were not able to account for differences in the relative severity of adverse events. For this reason, we presented the proportions of each of the individual adverse events. Tables 3–6 show that adverse events of various severity occur among both rural and referral patients in the facilities analysis, for both the 0 comorbidity level and the full patient risk strata. It should be noted that for appendectomy and hernia repair, there were high-severity adverse events that occurred only at referral facilities, albeit rarely, including pulmonary embolism, myocardial infarction, and bladder injury.
A 26% overlap in the ICD-10-CA codes used to identify adverse events with those used to assign Case Mix Group CL to patients causes a conceptual issue in using the comorbidity level to control for differences in the acuity of cases between patients in a regression analysis. For this reason, Case Mix Group CL was not used as a covariate in any regression models. Use of these comorbidity levels in risk stratification of our analysis was justified by inspection of Tables 3–6, which show whether patients with a comorbidity level of 0 still experience any of the individual adverse events.
In the analysis of outcomes by provider specialty, the proportion of patients in the FPESS and specialist groups were not equivalent and thus do not allow for a 2-way balanced design. This is not surprising given the low-procedural-volume niche that FPESS fill. However, we believe the volumes are high enough for a meaningful comparison.
Conclusion
We have shown evidence that the quality of care at rural facilities and, independently, care provided by FPESS is at least equivalent to that provided by referral facilities and specialists in rural BC, specifically for colonoscopy, hernia repair, and appendectomy. Likewise, although complications associated with cesarean delivery at RSON sites were 22% lower, volume was not adequate to achieve statistical significance. Further, we have presented a robust way of considering rural procedural outcomes in low-volume settings and within low-volume providers to ensure outcomes are commensurate to those of larger, specialist-led facilities.
Supplementary Information
Acknowledgements
The authors gratefully acknowledge funding for this work from British Columbia’s Joint Standing Committee on Rural Issues and support and collaboration from the Rural Coordination Centre of BC. The authors are also grateful to University of British Columbia students Payal Parti and Molly Rutherford, who contributed to background literature and manuscript preparation.
Footnotes
Preliminary results of this work were presented at the Quality Forum Conference in Vancouver in May 2022.
Competing interests: Sean Ebert reports a role with Pain BC. No other competing interests were declared.
Contributors: Jude Kornelsen led all aspects of this work and led manuscript writing. Gal Av-Gay provided expertise in conceptualizing the methodological approach, analyzed data, and contributed to the initial writing of the paper. Anshu Parajulee contributed to the methodological direction and analysis of the data. Nancy Humber contributed clinical expertise to the development of the indicators, analysis, and manuscript writing. Sean Ebert made contributions to the interpretation and analysis of the data and editing the final document. Tom Skinner made contributions to the design of the study and the acquisition, analysis, and interpretation of data. He participated in revising the work critically. Kathrin Stoll contributed to developing an analysis plan, and helped with writing, and critically revising the manuscript. All of the authors gave final approval of the version to be published and agreed to be accountable for all aspects of the work.
Population Data BC disclaimer: Access to data provided by the data steward is subject to approval but can be requested for research projects through the data steward or their designated service providers. All inferences, opinions, and conclusions drawn in this publication are those of the authors and do not reflect the opinions or policies of the data steward.
Funding: The Joint Standing Committee on Rural Issues provided funding (R005415). The funder had no role in the project.
References
- 1.Iglesias S, Kornelsen J, Woollard R, et al. Joint position paper on rural surgery and operative delivery. Can J Rural Med 2015;20:129–38. [PubMed] [Google Scholar]
- 2.Kornelsen J, Iglesias S, Woollard R. Sustaining rural maternity and surgical care: lessons learned. Can Fam Physician 2016;62:21–3. [PMC free article] [PubMed] [Google Scholar]
- 3.Kornelsen J, Khowaja A, Av-Gay G, et al. The rural tax: comprehensive out-of-pocket costs associated with patient travel in British Columbia. BMC Health Serv Res 2021;21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Iglesias S, Kornelsen J. An evidence-based program for rural surgical and obstetrical networks. Rural Remote Health 2018;18:4921. [DOI] [PubMed] [Google Scholar]
- 5.Health Authorities. Vancouver: Government of British Columbia. Available: https://www2.gov.bc.ca/gov/content/health/about-bc-s-health-care-system/partners/health-authorities (accessed 2023 Jan. 6). [Google Scholar]
- 6.Canadian Institute for Health Information. Discharge Abstract Database (Hospital Separations). V2. Vancouver: Population Data BC; 2022. Data Extract. MOH. Available: http://www.popdata.bc.ca/data (accessed 2025 May 18). [Google Scholar]
- 7.Perinatal Data Registry. Vancouver: Perinatal Services BC; 2022. Available: http://www.perinatalservicesbc.ca/health-professionals/data-surveillance/perinatal-data-registry (accessed 2025 May 18). [Google Scholar]
- 8.British Columbia Ministry of Health. Consolidation File (MSP Registration & Premium Billing). V2. Vancouver: Population Data BC; 2022. Data Extract. MOH. Available: http://www.popdata.bc.ca/data (accessed 2025 May 18). [Google Scholar]
- 9.Schuurman N, Fiedler RS, Grzybowski SC, et al. Defining rational hospital catchments for non-urban areas based on travel-time. Int J Health Geogr 2006;5:43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Av-Gay G, Stoll K, Parajulee A, et al. Evaluating health outcomes in rural British Columbia: a methodological approach using tests of non-inferiority and population-level data. Res Methods Med Health Sci 2023. [Google Scholar]
- 11.Hospital Harm Indicator: General Methodology Notes. Ottawa: CIHI; 2021. Available: https://www.cihi.ca/sites/default/files/document/hospital-harm-indicator-general-methodology-notes.pdf (accessed 2025 May 18). [Google Scholar]
- 12.Kelly C, Hulme C, Farragher T, et al. Are differences in travel time or distance to healthcare for adults in global north countries associated with an impact on health outcomes? A systematic review. BMJ Open 2016;6:e013059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Logroscino G, Marin B, Piccininni M, et al. Referral bias in ALS epidemiological studies. PLoS One 2018;13:e0195821. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Introduction to CMG+ Aggregation Variables. Ottawa: CIHI; 2022. Available: https://www.cihi.ca/sites/default/files/document/introduction-to-CMG-plus-aggregation-variables-en.pdf (accessed 2025 May 18). [Google Scholar]
- 15.Puhr R, Heinze G, Nold M, et al. Firth’s logistic regression with rare events: Accurate effect estimates and predictions? Stat Med 2017;36:2302–17. [DOI] [PubMed] [Google Scholar]
- 16.Mansournia MA, Geroldinger A, Greenland S, et al. Separation in logistic regression: causes, consequences, and control. Am J Epidemiol 2018;187:864–70. [DOI] [PubMed] [Google Scholar]
- 17.Rothmann MD, Wiens BL, Chan IS. Design and analysis of non-inferiority trials. CRC press; 2011. Jul 12:454. [Google Scholar]
- 18.Althunian TA, de Boer A, Groenwold RH, et al. Defining the noninferiority margin and analysing noninferiority: an overview. Br J Clin Pharmacol 2017;83:1636–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Tsong Y, Zhang J, Levenson M. Choice of δ noninferiority margin and dependency of the noninferiority trials. J Biopharm Stat 2007;17:279–88. [DOI] [PubMed] [Google Scholar]
- 20.VanderWeele TJ, Ding P. Sensitivity analysis in observational research: introducing the E-value. Ann Intern Med 2017;167:268–74. [DOI] [PubMed] [Google Scholar]
- 21.Reeve C, Humphreys J, Wakerman J. A comprehensive health service evaluation and monitoring framework. Eval Program Plann 2015;53:91–8. [DOI] [PubMed] [Google Scholar]
- 22.Urbach DR, Croxford R, MacCallum NL, et al. How are volume-outcome associations related to models of health care funding and delivery? A comparison of the United States and Canada. World J Surg 2005;29:1230–3. [DOI] [PubMed] [Google Scholar]
- 23.Moscovice I, Johnson K, Burstin H. Performance measurement in rural communities: the low-volume, large measurement challenge. Jt Comm J Qual Patient Saf 2017;43:259–62. [DOI] [PubMed] [Google Scholar]
- 24.Napier T, Olson JT, Windmiller J, et al. A long-term follow-up of a single rural surgeon’s experience with laparoscopic inguinal hernia repair. WMJ 2008;107:136–9. [PubMed] [Google Scholar]
- 25.Evans DV, Cole AM, Norris TE. Colonoscopy in rural communities: a systematic review of the frequency and quality. Rural Remote Health 2015;15:3057. [PubMed] [Google Scholar]
- 26.Azzopardi J, DeWitt DE. Quality and safety issues in procedural rural practice: a prospective evaluation of current quality and safety guidelines in 3000 colonoscopies. Rural Remote Health 2012;12:1949. [PubMed] [Google Scholar]
- 27.Edwards JK, Norris TE. Colonoscopy in rural communities: Can family physicians perform the procedure with safe and efficacious results? J Am Board Fam Pract 2004;17:353–8. [DOI] [PubMed] [Google Scholar]
- 28.Newman RJ, Nichols DB, Cummings DM. Outpatient colonoscopy by rural family physicians. Ann Fam Med 2005;3:122–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Iglesias S, Saunders LD, Tracy N, et al. Appendectomies in rural hospitals. Safe whether performed by specialist or GP surgeons. Can Fam Physician 2003;49:328–33. [PMC free article] [PubMed] [Google Scholar]
- 30.Lisonkova S, Sheps SB, Janssen PA, et al. Birth outcomes among older mothers in rural versus urban areas: a residence-based approach. J Rural Health 2011;27:211–9. [DOI] [PubMed] [Google Scholar]
- 31.McAteer JP, Richards MK, Stergachis A, et al. Influence of hospital and patient location on early postoperative outcomes after appendectomy and pyloromyotomy. J Pediatr Surg 2015;50:1549–55. [DOI] [PubMed] [Google Scholar]
- 32.Tom CM, Howell EC, Won RP, et al. Assessing outcomes and costs of appendectomies performed at rural hospitals. Am J Surg 2019;217:1102–6. [DOI] [PubMed] [Google Scholar]
- 33.Brajcich BC, Yang AD, Keswani RN, et al. The quality of screening colonoscopy in rural and underserved areas. Surg Endosc 2022;36:4845–53. [DOI] [PubMed] [Google Scholar]
- 34.Komaravolu SS, Kim JJ, Singh S, et al. Colonoscopy utilization in rural areas by general surgeons: an analysis of the National Ambulatory Medical Care Survey. Am J Surg 2019;218:281–7. [DOI] [PubMed] [Google Scholar]
- 35.Drukker L, Hants Y, Farkash R, et al. Impact of surgeon annual volume on short-term maternal outcome in cesarean delivery. Am J Obstet Gynecol 2016;215:85.e1–8. [DOI] [PubMed] [Google Scholar]
- 36.O’Connell RM, Abd Elwahab S, Mealy K. The impact of hospital grade, hospital-volume, and surgeon-volume on outcomes for adults undergoing appendicectomy. Surgeon 2020;18:280–6. [DOI] [PubMed] [Google Scholar]
- 37.Wei P-L, Liu S-P, Keller JJ, et al. Volume-outcome relation for acute appendicitis: evidence from a nationwide population-based study. PLoS One 2012;7:e52539. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.AlJamal YN, Zendejas B, Gas BL, et al. Annual surgeon volume and patient outcomes following laparoscopic totally extraperitoneal inguinal hernia repairs. J Laparoendosc Adv Surg Tech A 2016;26:92–8. [DOI] [PubMed] [Google Scholar]
- 39.Aquina CT, Kelly KN, Probst CP, et al. Surgeon volume plays a significant role in outcomes and cost following open incisional hernia repair. J Gastrointest Surg 2015;19:100–10. [DOI] [PubMed] [Google Scholar]
- 40.Aquina CT, Probst CP, Kelly KN, et al. The pitfalls of inguinal herniorrhaphy: surgeon volume matters. Surgery 2015;158:736–46. [DOI] [PubMed] [Google Scholar]
- 41.Aquina CT, Fleming FJ, Becerra AZ, et al. Explaining variation in ventral and inguinal hernia repair outcomes: a population-based analysis. Surgery 2017;162:628–39. [DOI] [PubMed] [Google Scholar]
- 42.Christophersen C, Baker JJ, Fonnes S, et al. Lower reoperation rates after open and laparoscopic groin hernia repair when performed by high-volume surgeons: a nationwide register-based study. Hernia 2021;25:1189–97. [DOI] [PubMed] [Google Scholar]
- 43.Christophersen C, Fonnes S, Baker JJ, et al. Surgeon volume and risk of reoperation after laparoscopic primary ventral hernia repair: a nationwide register-based study. J Am Coll Surg 2021;233:346–56. [DOI] [PubMed] [Google Scholar]
- 44.Christophersen C, Fonnes S, Andresen K, et al. Lower recurrence rate after groin and primary ventral hernia repair performed by high-volume surgeons: a systematic review. Hernia 2022;26:29–37. [DOI] [PubMed] [Google Scholar]
- 45.Köckerling F, Bittner R, Kraft B, et al. Does surgeon volume matter in the outcome of endoscopic inguinal hernia repair? Surg Endosc 2017;31:573–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Nordin P. Linden Wvd. Volume of procedures and risk of recurrence after repair of groin hernia: national register study. BMJ 2008;336:934. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Pace D, Borgaonkar M, Evans B, et al. Annual colonoscopy volume and maintenance of competency for surgeons. Surg Endosc 2017;31:2630–5. [DOI] [PubMed] [Google Scholar]
- 48.Baxter NN, Sutradhar R, Forbes SS, et al. Analysis of administrative data finds endoscopist quality measures associated with postcolonoscopy colorectal cancer. Gastroenterology 2011;140:65–72. [DOI] [PubMed] [Google Scholar]
- 49.Singh H, Penfold RB, DeCoster C, et al. Colonoscopy and its adverse events across a Canadian regional health authority. Gastrointest Endosc 2009;69:665–71. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.