Emergency general surgery: impact of distance and rurality on mortality

Abstract Background There is debate about whether the distance from hospital, or rurality, impacts outcomes in patients admitted under emergency general surgery (EGS). The aim of this study was to determine whether distance from hospital, or rurality, affects the mortality of emergency surgical patients admitted in Scotland. Methods This was a retrospective population-level cohort study, including all EGS patients in Scotland aged 16 years or older admitted between 1998 and 2018. A multiple logistic regression model was created with inpatient mortality as the dependent variable, and distance from hospital (in quartiles) as the independent variable of interest, adjusting for age, sex, co-morbidity, deprivation, admission origin, diagnosis category, operative category, and year of admission. A second multiple logistic regression model was created with a six-fold Scottish Urban Rural Classification (SURC) as the independent variable of interest. Subgroup analyses evaluated patients who required operations, emergency laparotomy, and inter-hospital transfer. Results Data included 1 572 196 EGS admissions. Those living in the farthest distance quartile from hospital had lower odds of mortality than those in the closest quartile (OR 0.829, 95 per cent c.i. 0.798 to 0.861). Patients from the most rural areas (SURC 6) had higher odds of survival than those from the most urban (SURC 1) areas (OR 0.800, 95 per cent c.i. 0.755 to 0.848). Subgroup analysis showed that these effects were not observed for patients who required emergency laparotomy or transfer. Conclusion EGS patients who live some distance from a hospital, or in rural areas, have lower odds of mortality, after adjusting for multiple covariates. Rural and distant patients undergoing emergency laparotomy have no survival advantage, and transferred patients have higher mortality.


Introduction
The impact of distance from hospital, and rurality, on mortality in emergency general surgery (EGS) patients remains unclear, with previous studies demonstrating a range of effects, from beneficial, to harmful [1][2][3][4][5][6][7][8][9] . Travel time to hospital was not a primary determinant of mortality in laparotomy audits in Britain or rural Australia 1,4 . Further studies demonstrated that it is safe to provide EGS laparotomies in non-urban centres in the USA and Australia 3,5,6 . In Scotland, one study showed distance was not related to mortality after ruptured abdominal aortic aneurysm 8 . A later Scottish study demonstrated decreased mortality with greater distance from the hospital but admitted the possibility of survival bias in their methodology 9 . In summary, the evidence is at best inconclusive, and at worst contradictory.
Many studies define EGS patients as those who have undergone an emergency operation. However, as less than 25 per cent of patients admitted under surgical services as an emergency undergo an operation, it is helpful to define EGS patients as all non-scheduled admissions under the care of a general surgeon 10 . It is also important to recognize that irrespective of whether patients live in an urban or rural setting, they may live very close to, or far away from an admitting EGS hospital. Therefore, it is useful to investigate both rurality and distance from hospital. It is not known whether patients who require an EGS admission are more likely to survive based on the distance from hospital or rurality. This question has profound implications for service delivery.
The aim of this study was to determine whether distance from hospital, or rurality, affects mortality of EGS patients in Scotland. Scotland has large remote and rural areas, particularly in the North and West of the country, and many islands. The hypothesis was that mortality increases in EGS patients as distance between home and hospital increases, and as rurality increases.

Data source
Administrative data from the Information Services Division of the Government of Scotland were routinely collected. This national database included population-level data of EGS patients during the study interval.

Population
An EGS patient was defined as a patient aged 16 years and older, non-electively admitted to a Scottish hospital, under the care of a consultant (attending) general surgeon for the full calendar years of 1998-2018 inclusive. Patients were followed up for 6 months.

Setting
Scotland has a national healthcare system where patients are treated at no direct cost to the patient. EGS care is provided by general surgeons working at teaching hospitals, large district general hospitals, and small district general hospitals 11 .

Data extracted
Data extracted included age at admission, sex, Charlson co-morbidity index (CCI, 10-year look-back), Scottish index of multiple deprivation (SIMD), admission origin (from home (domicile), transferred from another hospital, or other-including nursing homes, prisons, or no fixed abode), diagnosis (coded by use of the ICD-10) 12 , operations (coded by use of the OPCS-4) 13 , distance from hospital (calculated as the distance of a straight line between patient address and hospital address), year of admission, and date of death. SIMD is a measure of socioeconomic deprivation, comprehensively ranking all small geographical areas in Scotland (based on income, employment, education, health, access to services, crime, and housing), and then further classifying them in quintiles, with 1 as the most deprived and 5 as the least deprived 14 .

Analysis
The data were analysed with two logistic regression models. The first model explored the effect of distance to hospital, adjusting for variables chosen a priori that were previously shown to significantly affect the outcome of interest (mortality) 19,20 . The model defined inpatient mortality as the dependent variable, and distance from hospital as the independent variable of interest, adjusting for age, sex, co-morbidity (CCI), deprivation (SIMD), admission origin, diagnosis category, operative category, and year of admission. An identical model was created with 1-year mortality as the dependent variable. The second model explored the impact of rurality, with the six-fold SURC as the independent variable of interest. An identical model with 1-year mortality as the dependent variable was also analysed. The SURC incorporates several factors, and 'provides a consistent way of defining urban and rural areas across Scotland' 23 . Several versions are available, with varying numbers of categories, ranging from two to eight. The latest (eight-fold) version was published in 2016. Previous research has shown that within the eight-fold classification, category 5 residents have very short travel times to hospitals 24 . Therefore, the six-fold classification was chosen as the most detailed SURC available without inconsistency † Number is less than or equal to 5, or in the same row of a number that is less than or equal to 5, which may be identified.
regarding travel times (Table S2 and Fig. 1). Sensitivity analyses were repeated with the three-fold SURC, and two-fold SURC. Analyses were conducted with SPSS ® version 27 (IBM, Armonk, New York, USA).

Subgroup analyses
Analyses were repeated for several predefined subgroups, including patients who underwent operative treatment (Tables S3 and S4), patients who required emergency laparotomy (Tables S5 and S6), and patients who were transferred to a higher level of care (Tables S7 and S8).

Study conduct
This study was approved via the Public Benefit and Privacy Panel for Health and Social Care (PBPP 1819-0340) and did not require further research ethics approval. The STROBE guidelines were used to inform manuscript preparation (Table S9) 25 .

Results
There were 1 631 198 patients admitted to emergency general surgical services during the study interval. We excluded 48 351 who were aged under 16 years, 14 with missing sex data, and 10 637 whose place of residence could not be assigned to an SURC category, leaving a total of 1 572 196 admissions. Table 1 outlines the baseline characteristics of the population cohort by distance from hospital. We found that more young people lived closer to their admitting hospital than older patients. Sex proportions were similar across distance quartiles. A higher proportion of patients with moderate and severe co-morbidity lived further from their admitting hospital. Patients from deprived geographical areas (SIMD 1) were more likely to live close to their admitting hospital, whereas more of those in medium levels of deprivation (SIMD 3 and 4) lived further away. More patients were transferred who lived further away. A higher proportion of patients with high-risk diagnoses lived distant from their admitting hospital. Non-operative treatment was similar by distance quartile, but more laparotomies were performed for patients living further away. Table 2 outlines the baseline characteristics by six-fold SURC. There were many more patients who lived in urban geographical areas (SURC 1 and 2) than rural locations (SURC 3-6) ( Table 2). There were similar proportions of age group, sex, co-morbidity, origin, high-risk diagnosis, and treatment categories by SURC category; however, patients living in high levels of deprivation (SIMD 1 and 2) tended to live in urban areas (SURC 1 and 2), whereas a higher proportion of people living in medium deprivation regions (SIMD 3 and 4) lived in more-rural areas ( Table 2).
Table S10 displays rates of inpatient mortality based on distance from hospital (in quartiles), by age category, sex, co-morbidity, deprivation, origin, diagnosis categories, treatment categories, and six-fold SURC. A higher proportion of patients died who lived closer to the admitting hospital, who were older, female, highly co-morbid, admitted from  (7) *Cannot calculate percentage. † Number is less than or equal to 5, or in the same row of a number that is less than or equal to 5, which may be identified. SURC, Scottish Urban Rural Classification; CCI, Charlson co-morbidity index; SIMD, Scottish index of multiple deprivation (1, most deprived; 5, least deprived); GI, gastrointestinal.
non-domicile environments, with high-risk diagnosis, and who had a laparotomy. Table S11 demonstrates inpatient mortality for each six-fold SURC category.

Distance from hospital and mortality
Those admissions of patients who lived in the furthest quartile from hospital had lower odds of mortality than those in the closest quartile (OR 0.829, 95 per cent c.i. 0.798 to 0.861), although there was no statistically significant difference between the first and second, or first and third quartiles (

Rurality and mortality
Patients who lived in the most rural category (SURC 6) had higher odds of survival than those in the most urban category (SURC 1) (OR 0.800, 95 per cent c.i. 0.755 to 0.848) (

Patients who required emergency laparotomy
Subgroup analysis, including only those who underwent an emergency laparotomy (n = 20 669), showed no significant difference of inpatient or 1-year mortality either in distance from hospital (Table S5), or rurality (Table S6).

Patients who were transferred
Analysis of those who were transferred between hospitals (n = 11 869) showed increased mortality with increased distance from hospital and increased rurality (Tables S7 and S8). There were significant increases of inpatient mortality in the third distance quartile (OR 1.520, 95 per cent c.i. 1.019 to 2.266), and 1-year

Discussion
For EGS patients, including those who were managed non-operatively, there seems to be increased survival for those residing further from the admitting hospital, and/or in more remote locations. This paradoxical finding may be explained by the types of patients who are admitted under general surgical care in the UK, many of whom suffer from low-acuity conditions. This beneficial effect is no longer apparent when patients with more serious illness-such as those requiring emergency laparotomy-are considered, or those patients who are transferred between hospitals because they require a high level of care. However, it is reassuring to know that patients who reside in remote and rural areas and require emergency laparotomy do not have worse mortality than those who live in more central locations.
The evidence from similar published literature regarding whether rurality or distance from hospital affects mortality is inconsistent, ranging from beneficial to detrimental. A UK study that evaluated patients included in the National Emergency Laparotomy Audit from 2013-2016 showed that the estimated travel time between home and hospital was not a primary determinant of short-term mortality 1 . The rural emergency laparotomy audit in Australia reported similar findings 4 . In the USA, the occurrence of adverse postoperative events after EGS was not related to whether patients lived in rural areas 3 . Other studies confirm the safety of undertaking emergency abdominal surgery in non-urban centres 5,6 . A Scottish study found that distance from hospital had no significant impact on community mortality rates of ruptured abdominal aortic aneurysms 8 . More recently a study demonstrated that increased distance to hospital led to decreased mortality after open repair of ruptured abdominal aortic aneurysm between 1990-2011; however, the authors suggested that this could have been due to survivor bias 9 . In fields related to EGS, the evidence is similarly inconsistent. In trauma care, several US studies have shown increased mortality risk for rural trauma populations [26][27][28][29] . However, a large US study found that mortality did not differ between rural and urban regions, even though a higher proportion of rural deaths occurred within 24 h compared with urban deaths 30 . In Scotland, long prehospital times in rural environments did not affect mortality in moderately and severely injured patients 31 .
There are several possible explanations for the apparent benefit seen in the study population. First, patients who did not survive to hospital admission were not included. It is possible that a greater proportion of patients from longer distances, or   Similarly, surgeons receiving referrals from rural or longer distance areas, may be more willing to accept the patient under their care, because of the lack of alternative resource and the medical consequences of not addressing major pathology that presents with mild symptoms. Third, there may exist unknown confounders that provide survival benefit to those living in rural locations, or locations far from major hospitals. Perhaps certain lifestyle factors specific to rural areas better prepare patients for EGS admissions, which improves their odds of survival, such as levels of physical activity [35][36][37] . There is a complex relationship between rurality, socioeconomic status, and physical activity, such that as remoteness and socioeconomic status increased, physical activity increased in Australia 38 . Although adjusted for in modelling, the data in this study demonstrate that a high proportion of Scottish urban dwellers (SURC 1 and 2) live in low-deprivation areas (SIMD 1 and 2), whereas a comparatively small proportion of rural dwellers (SURC 5 and 6) live in low-deprivation areas ( Table 2). These data demonstrating that rural survival is higher may be because rural dwellers represent a less-deprived cohort, which cannot be ruled out completely. This study has limitations. There are several variables that are not accounted for in the data, which would have provided a more thorough risk adjustment, relating to physiological, or biochemical parameters. Transfer information did not include the specific hospital or level of hospital that patients were transferred from, precluding a more detailed analysis. Both rurality and distance from hospital were analysed, which are related but different. For example, a patient living in a rural location may live very close to their admitting hospital (rural, but short travel time); however, such hospitals are usually smaller with fewer facilities. Conversely, a patient may be living in a less rural location, but their closest admitting hospital may be 20 km away (not rural, but long travel time). Adjusting for hospital type and volume was considered, but there was collinearity between these variables and rurality, so they were excluded from the analyses; however, a recently published paper addressed the impact of hospital and surgeon admission volume, and mortality 39 . If distance from hospital was associated with outcome, one would expect this association to be linear, but those in the third quartile did not have better survival. This may reflect other hidden confounders. Finally, the event (mortality) rates were very low (1-2 per cent), which would normally affect the performance of logistic regression models, but because of the large sample size the analyses are still valid.
This study also has many strengths, the most important being its size, and the quality, and consistency of the data. Scotland's population-based health data are regularly audited for accuracy and include a validated urban/rural classification that facilitates studies of this kind 40 . These findings have important health policy implications. Rural surgery is an important part of healthcare provision in geographically dispersed populations [41][42][43] . Access to surgical services, especially out of hours, can be limited for those in remote and rural areas [44][45][46] . This is becoming increasingly problematic in some areas, including parts of the USA, where the rate of rural hospital closures has risen in the past decade, largely due to financial, market, and staffing issues 45 . Centralization of health services is an obvious but contentious solution, given the cost and inconvenience of travelling large distances from patients' home to receive care 47 , and the results do not support this strategy. The Scottish Government published a report stating the role of rural general hospitals for EGS patients: '24-hour surgical services should provide local assessment, triage, resuscitation stabilization of emergency surgical and trauma patients followed by admission and surgical intervention, if appropriate, and transfer, when necessary, in collaboration with the relevant receiving hospital' 48 . Clearly, this paper addresses a topic of great importance to the Scottish health system, for which policies and procedures have been intentionally designed to avoid further centralization by providing adequate rural care and allowing transfer when necessary.
A key area of future research is the early identification and prognostication of patients at high risk of requiring transfer to higher levels of care from rural and distant populations. Several clinical decision support tools that predict mortality and need for intensive care have been developed 49,50 . Trauma systems have widely adopted trauma field triage decision tools to decide whether to bypass smaller trauma units and convey to large trauma centres 51,52 . An analogous system devised for EGS patients who may require transfer to centres with specialist surgical services or ICUs, may improve care for rural and distant populations 53 .

Funding
This work was made possible by a grant from NHS Grampian and NHS Highlands Endowment Funding. No funding was received from the National Institutes of Health (NIH); Wellcome Trust; or Howard Hughes Medical Institute (HHMI).