Secondary Logo

Journal Logo

Original Research Reports: Original Clinical Research Report

Changes in Surgical Volume and Outcomes During the Coronavirus Disease 2019 Pandemic at Two Tertiary Hospitals in Ethiopia: A Retrospective Cohort Study

Alshibli, Amany K. BS*; Teklehaimanot, Masresha G. BSc; Seyoum, Rahel MSc; Tegu, Gebrehiwot A. MSc; Desta, Haftom B. MSc; Gong, Wu MS, MD§; Tesfaye, Gosa MD; Tsegaw, Agenchew N. MSc; Abay, Abraha Y. BSc; Etanaa, Naod B. MSc; Mossie, Mulat MSc; Benwu, Kore M. MSc; Tarekegn, Fantahun MSc; Gebremedhin, Hagos G. MSc; McEvoy, Matthew D. MD; Newton, Mark W. MD; Sileshi, Bantayehu MD

Author Information
doi: 10.1213/ANE.0000000000005946

Abstract

KEY POINTS

  • Question: Is there an association between the coronavirus disease 2019 (COVID-19) pandemic and the provision of surgical care and perioperative outcomes at 2 low-resource referral hospitals in Ethiopia?
  • Findings: We observed a reduction of surgical case volume and a shift in referral patterns during the pandemic; however, our analyses did not show a significant increase in perioperative mortality for surgeries performed during the COVID-19 pandemic.
  • Meaning: Patients may delay seeking or reaching care during the pandemic, but safe surgery was maintained despite COVID-19-related infection control restrictions.

An estimated additional 143 million surgical cases per year are needed to meet the burden of surgical disease in low- and middle-income countries (LMICs).1–3 A global effort to build surgical capacity in LMICs is ongoing under working goals outlined by the Lancet Commission on Global Surgery.1 Many of these efforts were underway when the novel coronavirus severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was declared a global pandemic.4

Disruption to routine health care services has been multifaceted since December 2019, when the disease caused by SARS-CoV-2 (coronavirus disease 2019 [COVID-19]) was first identified.5 The “3-delays” model—delays in seeking, reaching, and receiving care—provides a framework to think about the social, geographic, and economic factors impacting patients with surgical conditions during this time.6 There has been an unprecedented worldwide cancellation of elective and nonurgent surgeries in an effort to curb infection, ration resources, and create hospital COVID-19 surge capacity.7 One model estimated a 72% surgery cancellation rate during the peak 12-week disruption of the COVID-19 pandemic.7 Beyond the immediate impact of COVID-19 infection, experts fear the “collateral damage” resulting from delaying care for unrelated conditions.8,9 Early reports in the literature, largely from high-income countries (HICs), are suggesting delays in care leading to poor outcomes. A single-center study in Italy reported an increase in perioperative complication rate and a 41.3% decline in emergency surgeries performed during March and April 2020 compared to the same period in 2019.10 Some sites report decreased incidence of admissions for acute appendicitis11–13 and acute myocardial infarction14–16 during the peak of the COVID-19 pandemic.14 One hypothesis to explain these findings is that patients are delaying presenting for care for fear of viral transmission at health care facilities.11,14,17

Little is known about how this pandemic affects the provision of surgical care and perioperative outcomes, especially in a low-resource setting. We sought to describe the association between the start of the COVID-19 pandemic and both surgical care parameters and outcomes at 2 tertiary hospitals in Ethiopia using perioperative data collected with a previously validated electronic data collection tool.18 The main study objectives were to: (1) describe surgical case volume before and during the pandemic, (2) describe referral patterns before and during the pandemic, and (3) report odds of perioperative mortality adjusted for confounders before and during the pandemic.

METHODS

Study Design

After institutional review board (IRB) approval, including informed consent waiver by all IRBs (IRBs: Ayder Comprehensive Specialized Hospital [ACSH], Mekelle, Ethiopia, Africa; Tibebe Ghion Specialized Hospital [TGSH], Bahir Dar, Ethiopia, Africa; and Vanderbilt University Medical Center [VUMC], Nashville, Tennessee, USA), we conducted a retrospective observational cohort study comparing surgical cases performed before and during the COVID-19 pandemic at 2 tertiary hospitals in Ethiopia. ACSH in Mekelle, Ethiopia, is a tertiary care academic hospital associated with Mekelle University that provides specialty care to the Tigray region. TGSH in Bahir Dar, Ethiopia, is similarly a tertiary care academic hospital associated with Bahir Dar University in the Amhara region. The Ethiopian Ministry of Health identified these sites as academic training hubs suited for a capacity-building partnership with VUMC in Nashville, Tennessee. This study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.

We defined “surgical cases” as all procedures performed in an operating room under the supervision of an anesthesia provider for the specified time periods. We defined 3 “phases” or cohorts of the pandemic exposure a priori. We identified phase 0, the control cohort, as surgical cases performed during the selected prepandemic time period: August–September 2019 for TGSH and November–December 2019 at ACSH. We chose these dates individually for each hospital to minimize the bias of external factors unrelated to the pandemic, such as power outages and holidays, affecting surgical volume. We defined phase 1 as surgical cases performed during the period of time at which the hospital instituted policies canceling all elective surgeries: April 1–June 10, 2020, and July 8–August 3, 2020, at ACSH and March 28–April 12, 2020, at TGSH. Surgical cases performed at ACSH in the month between June 10, 2020, and July 8, 2020, were excluded from the analysis since the hospital temporarily resumed elective surgery activity during that time. We chose this definition of the exposure because of reliability across settings, and because it was considered the best surrogate for when the concern for local COVID-19 cases was significant enough to warrant this policy. We defined phase 2 as surgical cases performed from the time hospital policies allowed elective surgeries to resume, August 4 at ACSH and April 13 at TGSH, until the end of the study on August 31, 2020. We included phase 2 in the analysis to understand changes in surgical care beyond the initial phase of the pandemic.

Questionnaire

The Research Electronic Data Capture (REDCap) tool that we used to prospectively collect data on surgical cases in LMICs has been previously described.18 It consists of 132 data elements that were selected based on clinical significance and factors that reportedly influence surgical outcomes (eg, safe surgery checklist use) while minimizing the length of provider time needed for completion (on average, collecting data on 1 patient takes <10 minutes). The tool comprises 5 sections: provider information and case demographics, general case information, preoperative vital signs, anesthetic information, and follow-up and complications. A nonclinical research staff member, referred to as a “data manager,” was hired at each hospital to assist with collecting follow-up data and maintaining the electronic database. The primary anesthesia provider or student entered initial data using electronic tablets, with the option for offline entry; data managers performed collection of follow-up data (including mortality status and discharge status up to 28 days postoperative) and completion of missing data. All data collectors, including data managers, the in-room anesthesia provider, and students who accessed the database, received training on how to use the electronic tool and on principles of data management and quality improvement. This included online modules and in-person simulation of data collection, as well as training on ethics and challenges of clinical data collection. Data managers routinely cross-referenced the captured data with hospital logbooks. Data were stored in a secure online REDCap database.

Regional Follow-up

Data managers who followed the postoperative course of all patients until 28 days postoperative obtained the mortality data. The data managers determined mortality status using either inhospital information or a follow-up phone call to the patient or designated caregiver. To find data missing from the REDCap database, data managers retrospectively collected information from hospital records and through follow-up for purposes of this study. We included all available cases recorded during the a priori specified study periods in our analysis, omitting any data points that were missing after retrospective collection due to incorrect follow-up information (eg, medical record number and phone number).

Statistical Analysis

We established primary outcomes at the initiation of study design, before data exploration. We created tables to describe the patient population at each hospital and in each phase cohort. We grouped data on patient district of origin into 3 categories (local district of the hospital, outside the local district, and outside the region) and tabulated the number and proportion of patients in each category for hospitals. We performed a complete case analysis and presented P values without familywise adjustment.

We estimated case volume by dividing daily case frequency in REDCap by monthly percent of total cases recorded in the database. We plotted boxplots to visualize the distribution of daily case volume for weekdays during each phase at hospitals. To visualize trends over time, we plotted LOESS curves with 95% confidence intervals (CIs) of unadjusted case volume and mean proportion of categorical variables’ emergency status and district of origin. LOESS curves were generated using the “ggplot2” package in R, with a span of 75% used. We estimated an incidence rate ratio (IRR) of daily case volume with a piecewise Poisson regression model controlling for day of the week and hospital. We segmented the data with time represented as a categorical variable with 4 levels (phase 0, time period between phases 0 and 1, phase 1, and phase 2).

Table. - Demographics of Patients Undergoing Surgery During 3 Phases
Phase 0 Phase 1 Phase 2
November 1–December 31, 2019, ACSH April 1–August 3, 2020, ACSH August 4–August 31, 2020, ACSH
August 1–September 30, 2019, TGSH March 28–April 12, 2020, TGSH April 13–August 31, 2020, TGSH
Number of cases (n) 1173 558 1500
Age 28.00 (19.00–40.00) 28.00 (21.00–38.00) 28.00 (20.00–43.00)
Sex
 Female 529 (47.4) 262 (49.4) 711 (47.5)
 Male 586 (52.6) 268 (50.6) 785 (52.5)
Urgency
 Elective 607 (53.2) 126 (22.7) 698 (46.9)
 Emergency 534 (46.8) 429 (77.3) 789 (53.1)
Trauma 258 (22.6) 154 (27.8) 342 (22.9)
ASA status
 I 655 (56.3) 125 (22.6) 763 (51.0)
 II 438 (37.7) 364 (65.8) 643 (43.0)
 III 70 (6.0) 64 (11.6) 89 (6.0)
Comorbidity
 Anemia 30 (2.6) 10 (1.8) 41 (2.7)
 DM 14 (1.2) 6 (1.1) 25 (1.7)
 HIV 24 (2.0) 5 (0.9) 20 (1.3)
 HTN 52 (4.4) 24 (4.3) 98 (6.5)
 Other 90 (7.7) 23 (4.1) 209 (13.9)
 None 990 (84.4) 503 (90.1) 1159 (77.3)
Time of surgery
 Daytime 730 (63.9) 379 (69.0) 1083 (72.9)
 Nighttime 355 (31.1) 155 (28.2) 377 (25.4)
 Weekend 57 (5.0) 15 (2.7) 26 (1.7)
Procedure group
 Cesarean delivery 141 (12.3) 106 (19.0) 238 (15.9)
 General surgery 233 (20.2) 114 (20.5) 420 (28.1)
 Orthopedic 307 (26.7) 125 (22.4) 334 (22.4)
 Othera 470 (40.8) 212 (38.1) 502 (33.6)
Surgery length (min) 60.00 (40.00–120.00) 70.00 (50.00–100.00) 70.00 (45.00–120.00)
Anesthesia type
 General 586 (51.0) 285 (51.4) 778 (52.6)
 Regional 502 (43.7) 262 (47.3) 635 (42.6)
 Other 61 (5.3) 7 (1.3) 68 (4.6)
Safe surgery checklist used?
 Yes 1064 (92.4) 535 (96.1) 1491 (99.7)
 No 87 (7.6) 22 (3.9) 4 (0.3)
Data are presented as absolute counts (%) for categorical and median (IQR) for continuous variables age and surgery length.
Phase 0 = prepandemic, phase 1 = lockdown, and phase 2 = postlockdown.
Abbreviations: ACSH, Ayder Comprehensive Specialized Hospital; ASA, American Society of Anesthesiologists; DM, diabetes mellitus; ENT, ear, nose, and throat; HIV, human immunodeficiency virus; HTN, hypertension; IQR, interquartile range; TGSH, Tibebe Ghion Specialized Hospital.
a“Other” procedure group includes subspecialty surgeries: ENT, neurosurgery, cardiothoracic, urology, and ophthalmology.

F1
Figure 1.:
Flow diagram of study sample; we excluded cases from an intermediate period in phase 1 at ACSH during which the hospital resumed elective surgery activity for a brief period before cancelling elective surgeries again. All available data from 3231 cases were included in the analysis of case volume, referral patterns, and demographics. Cases missing mortality data were excluded from the logistic regression model. ACSH indicates Ayder Comprehensive Specialized Hospital.
F2
Figure 2.:
Distribution of estimated weekday case volume for (A) all cases, (B) emergency cases, and (C) elective cases. Estimates were obtained by adjusting the weekday number of cases recorded in the REDCap database by the monthly proportion of logbook cases recorded in the database. ACSH indicates Ayder Comprehensive Specialized Hospital; REDCap, Research Electronic Data Capture; TGSH, Tibebe Ghion Specialized Hospital.
F3
Figure 3.:
Region or district of origin of patient undergoing surgery at each hospital, presented as % of cases. ACSH is located in Mekelle in the Tigray region, and TGSH is located in Bahir Dar in the West Gojjam Zone of the Amhara region. Phase 0 = prepandemic, phase 1 = lockdown, and phase 2 = postlockdown. Map data from R. Hijmans and University of California, Berkeley, Museum of Vertebrate Zoology (2015). Second-level Administrative Divisions, Ethiopia, 2015. UC Berkeley, Museum of Vertebrate Zoology. Available at: http://purl.stanford.edu/ny194cj6506. ACSH indicates Ayder Comprehensive Specialized Hospital; TGSH, Tibebe Ghion Specialized Hospital.
F4
Figure 4.:
Multivariable logistic regression to describe association between 28-day perioperative mortality and phase of pandemic, adjusting for potential confounders. Phase 1 and phase 2 surgeries were not significantly associated with increased mortality when controlling for confounders. Other variables are included in model to adjust for confounding, and those associated with mortality include age, ASA physical status ≥2, emergency status, and surgery length. “Other procedure type” includes subspecialty surgeries: ENT, neurosurgery, cardiothoracic, urology, and ophthalmology. Phase 0 = prepandemic, phase 1 = lockdown, and phase 2 = postlockdown. ASA, American Society of Anesthesiologists; CI, confidence interval; ENT, ear, nose and throat; OR, odds ratio; Ref, reference variable.

In modeling the time trend, to avoid unnecessary nonlinear assumptions, we simplified the time trend with a piecewise segmented linear model with slopes forced to 0. The coefficients of phases in this model gave us the direct estimate of the average change (ratio) of the incidence rate of daily cases. We found an interaction between emergency status and phase on the outcome, and so conducted subgroup analysis of elective and emergency surgeries. We used Poisson regression models to estimate the IRR of daily cases for emergency and elective surgeries. The Poisson regression model for subgroups was similar to the model for overall case volume and included time as a categorical variable with 4 levels, day of the week, and hospital. To describe association of the exposure variables with the outcome of 28-day mortality, we used logistic regression with covariates of age, American Society of Anesthesiologists (ASA) physical status, urgency, procedure type, surgery length, and anesthesia type. We chose these covariates primarily based on clinical significance or because they differed significantly by mortality status in univariate analysis (see Supplemental Digital Content 1, Table 1, https://links.lww.com/AA/D863). We plotted adjusted odds ratios (aORs) for the mentioned covariates and 95% CIs on a logarithmic scale. We used R software to perform all analyses.

RESULTS

We captured perioperative data from 1217 surgical cases from ACSH and 2014 cases from TGSH (N = 3231) in the REDCap database during the study period. We recorded 1173 cases in phase 0, 558 in phase 1, and 1500 in phase 2 (Figure 1). Overall, 299 cases (9.25%) were missing mortality outcome data and 24 (0.7%) were missing district of origin data, which could not be obtained retrospectively owing to incorrect or missing identifying information. We observed a similar patient population at both hospitals, with median age of 28 (interquartile range [IQR], 20–40) and 1502 (47.8%) female patients (see Supplemental Digital Content 2, Table 2, https://links.lww.com/AA/D864). The most common surgery types overall were orthopedic 766 (23.9%), general surgery 767 (24.0%), and cesarean delivery 485 (15.1%). During phase 0, 534 (47%) cases were emergency status, compared to 429 (77%) during phase 1 and 789 (53%) during phase 2 (see Table). There were 70 (6%) patients that were ASA physical status ≥3 during phase 0 and 64 (11.6%) during phase 1. Standardized mean differences for demographic data in phases 0–2 are shown in Supplemental Digital Content 3, Table 3, https://links.lww.com/AA/D865.

Case Volume

Overall, median estimated weekday case volume (cases/day) was 15.6 (IQR, 12.2–19.9) during phase 0, 7.2 (IQR, 3.2–12.9) during phase 1, and 11.9 (IQR, 6.8–17.2) during phase 2 (see Figure 2). In a Poisson regression model of daily case rate adjusted for hospital and day of the week, we observed a significant decrease in overall case volume during phases 1 and 2 compared to phase 0, with an IRR of 0.68 (95% CI, 0.63–0.73) for the time period between phases 0 and phase 1, 0.73 (95% CI, 0.66–0.81) for phase 1, and 0.90 (95% CI, 0.83–0.97) for phase 2 (see Supplemental Digital Content 4, Table 4, https://links.lww.com/AA/D866). A test for interaction of emergency status and phase on case volume was significant (P value <.001), so we conducted subgroup analysis. In the emergency case subgroup analysis, the volume of emergency case significantly increased during phase 1 with an adjusted IRR of 1.36 (95% CI, 1.19–1.55) compared to phase 0. However, the volume of emergency cases did not show a significant change in phase 2 compared to phase 0 with an adjusted IRR of 0.97 (95% CI, 0.87–1.09). For the elective case subgroup, we observed a significant decrease in elective case volume during phases 1 and 2 with an adjusted IRR of 0.29 (95% CI, 0.24–0.35) and 0.86 (95% CI, 0.77–0.97), respectively, compared to phase 0 (see Supplemental Digital Content 4, Table 4, https://links.lww.com/AA/D866 for subgroup analysis).

Referral Patterns

The proportion of patients from outside the district undergoing surgery shows different patterns during the pandemic for each hospital. At ACSH, 193 (39.6%) patients were from outside Mekelle during phase 0 compared to 248 (49.1%) and 115 (55%) during phases 1 and 2, respectively. At TGSH, 257 (38.5%) patients were from outside the district during phase 0, 11 (22%) during phase 1, and 416 (32.3%) during phase 2 (see Figure 3). Very few patients (N = 18% and 0.56%) were from outside the country throughout the study period. aOR of a patient coming from an outside district during lockdown was estimated using a logistic regression model with covariates age, sex, and procedure group. Phase 0 was the reference group to compare adjusted odds for phases 1 and 2. At ACSH, there were more patients from outside the district during phases 1 and 2 compared to phase 0 with aOR 1.63 (95% CI, 1.24–2.15) and 2.03 (95% CI, 1.43–2.90), respectively. While at TGSH, there was a significant decrease in patients from outside the district with an aOR of 0.44 (95% CI, 0.21–0.87) in phase 1 and 0.80 (95% CI, 0.65–0.99) in phase 2 compared to phase 0 (see Supplemental Digital Content 5, Table 5, https://links.lww.com/AA/D867).

Perioperative Mortality

There were 77 (2.63%) total mortalities recorded in the study population: 19 (1.8%; 95% CI, 1.2–2.8) in phase 0, 17 (3.7%; 95% CI, 2.3–5.8) in phase 1, and 41 (2.9%; 95% CI, 2.1–3.9) in phase 2. Compared with phase 0, unadjusted ORs of 28-day mortality were 2.07 (95% CI, 1.07–4.02) for phase 1 and 1.63 (95% CI, 0.94–2.83) for phase 2 (Figure 4). The aORs of 28-day mortality were 1.36 (95% CI, 0.62–2.98) for phase 1 and 1.54 (95% CI, 0.80–2.95) for phase 2, adjusting for age, ASA physical status, urgency, procedure type, surgery length, and anesthesia type.

Trends Over Time

To further examine the described results and observe the overall unadjusted trends, we plotted outcomes and descriptive variables over time (see Appendix, Figure A1). For case volume at ACSH, the LOESS curve is the highest before phase 0 and lowest around March 2020 just before the start of phase 1. At TGSH, we observed a lower case volume just before phase 0 and before the start of phase 1 in March 2020. We observe the highest proportion of emergency cases during phase 1 at ACSH. We observed a decrease for perioperative mortality rate (POMR) over time in phase 0 from both hospitals during August–December 2019, followed by a positive trend in part of phase 1 for hospitals until approximately May 2020. The proportion of patients coming from outside the hospital districts in phase 1 lockdown period showed opposite trends during phase 1.

DISCUSSION

We described surgeries before and during the COVID-19 pandemic. First, we observed a significant reduction in surgical case volume during and after COVID-related restrictions, primarily of elective cases. Second, we found that referral patterns shifted during the pandemic for each hospital. Third, we did not find a difference in odds of perioperative mortality comparing prepandemic and pandemic care. We discuss these findings in light of current literature.

Previous studies have reported a reduction in surgical case volume during the pandemic.7,13,19,20 We report similar findings in a setting where at baseline almost half of cases are emergency, which is unlike most high-resource centers.21 The reasons for this reduction may include the strict preventative measures that were necessary to keep COVID transmission low22 and patient fear of obtaining the virus. Because the incidence of most surgical conditions was likely unchanged during the pandemic, reduced surgical volume suggests delays in care that could contribute to poor outcomes, particularly for cancer surgery.22 A greater proportion of higher risk patients (ASA physical status ≥3) underwent surgery during phase 1. This could be explained by how hospitals defined essential surgeries, patient populations with more comorbidities, and patient delay in seeking care resulting in sicker patient at time of surgery. Public health campaigns should encourage patients to seek non-COVID-related care when necessary, especially when strict lockdown measures may decrease health-seeking behavior. Additionally, efforts to rapidly increase surgical capacity are necessary to address the “backlog” of delayed cases.23 One model estimates that this would require hospitals to operate at 120% capacity for 45 weeks,7 which would be near impossible without attention and resources to this issue.

We inferred referral patterns based on patient district of origin and found that trends differed between regions. At TGSH, there was a significant decrease in proportion of patients from outside the district during phase 1, suggesting that patients may have encountered difficulty traveling to the hospital due to restrictions. It is possible that other factors influenced this observation, given that the time plot in Figure A1 shows this trend began in January 2020 (see Appendix). Contrarily, at ACSH, the observed increase in proportion of patients from outside districts during phase 1 suggests that there could be a shift in the provision of surgical care from hospitals closer to their home to urban facilities. These patterns reveal that patients may experience delays in reaching surgical care that are not fully captured with hospital-level data. Postpandemic recovery of the health care system should incorporate public health strategies to better understand patient access limitations and allocate resources to maintain essential surgical care at the district hospital level.

The odds of mortality for patients undergoing surgery during the pandemic were not statistically significant, suggesting insufficient data to conclude that patients were more likely to experience poor outcomes during this period. We observed these results despite reports of alarmingly high rates of complication and mortality for patients with perioperative COVID-19 infection.17,20,24 We are unable to comment on complication rates since we only looked at mortality. Furthermore, we have not captured data on COVID-19 status in our cohort and are unable to sufficiently assess the effect of infection, if any, on outcomes observed. We believe that this would be insignificant given that these hospitals were not routinely admitting COVID-19 patients.22 We are unable to make a definitive negative conclusion about the hypothesis that mortality outcomes were unchanged for surgeries during the COVID-19 pandemic, given the wide CIs for odds ratio (OR) for mortality. There are studies reporting that mortality outcomes were unchanged for acute surgery in Spain20 and spine surgery in New York City.25 These studies occurred in high-resource settings experiencing COVID-19 surges, which limits this comparison. Overall, we provide evidence that essential surgical care may be maintained at a low-resource setting in the context of COVID-19-related restrictions.

While not the focus of this study, we demonstrate evidence for factors associated with perioperative mortality at a low-resource center. Age, ASA physical status ≥2 surgery length, and emergency status were associated with significantly increased odds of perioperative mortality, while regional anesthesia was associated with decreased odds of mortality, consistent with previous literature.18,21,26 The ORs reported should be interpreted with caution given that the logistic regression was modeling a rare outcome (ie, 77 mortalities). The 28-day POMR during the study period was 77 of 2932 (2.63%), which is likely equivalent to reporting standard of 30-day POMR. The follow-up period decided at the initial design of the data collection tool was 28 days for ease of routine weekly follow-up, before the 30-day POMR became standard. While it is challenging to compare POMR for different definitions and contexts,27 this finding is comparable to POMR findings from previous studies in Africa.18,21,28 Generalizations may not apply where patient population, procedure types, and resources differ.29

We believe that this is the first data-driven report assessing surgical care within East Africa during the pandemic. Collecting perioperative data prospectively—a strength of this study—is rare in low-resource settings, where electronic health records are usually limited. We addressed missing data and limitations in several ways. First, data managers retrospectively collected missing data. Records that had inadequate follow-up information were finally categorized as missing and were omitted. Previous validation of the data collection tool used demonstrated that outcomes did not differ significantly between observed and nonobserved cases,18 so we believe that this approach did not bias results. If there is potential bias of the absolute estimates due to missing data, such as for case volume, it is unlikely to affect comparisons between phases. Secondly, with human data entry, there is the possibility of response bias or error. We mitigated this by training providers on data entry principles, screening study data for errors, and cross-checking data with hospital logbooks. Finally, the COVID-19 pandemic is a complex exposure that varied between regions and over time. We considered using regional and hospital COVID-19 case numbers to measure the virus exposure, but due to variations in testing, we standardized the study design by defining exposure as the period during which hospital officials determined the local virus burden warranted elective surgery cancellation. Additionally, we chose the phase 0 period individually for each hospital to minimize bias caused by context factors unrelated to the pandemic that influenced surgical care, such as power outages and holidays. We have run a Poisson regression to model the incidence rate of case volume with limited adjustment for confounding. Case volume could be influenced by other facility-level or regional factors such as seasonality, infrastructure, and human resources. We were limited in our ability to fully adjust for all confounding variables without data on these factors, which means our results are likely subject to some degree of confounding bias. The observed OR of mortality comparing phases 0 and 1 was not statistically significant, and readers should be cautious in interpreting this result because this study had limited power to detect the mortality rate change. Given the observed sample size of 1173 cases in phase 0 and 558 cases in phase 1, given an expected mortality rate of 2.1% reported from a large surgical cohort in Africa,21 and using a 2-sided test with 5% type I error, we would have 7.0% power to detect a 25% relative mortality rate change, which is the effect size we consider as the minimum meaningful increase of mortality rate. We reported P values without familywise adjustment, and readers should be cautious in interpretating the significance of our findings.

During the COVID-19 pandemic, reduction of surgical volume and changes in referral patterns persisted even after restrictive policies were lifted. Provider discretion, hospital resources, and patient decisions to seek care may all have influenced the observed results. Health care systems must design and implement “surgical recovery plans” to clear the backlog of delayed cases and maintain the provision of essential surgical care, which is especially vital in a setting where the burden of surgical disease is already catastrophic. Future work can characterize complication rates and perioperative infection in this population and model scenarios for clearing the surgical backlog in this setting.

ACKNOWLEDGMENTS

The authors thank the Bahir Dar University and Mekelle University students, staff, and faculty involved in data collection; the ImPACT Africa team, data managers, staff, and providers for their hard work and vision; and the Vanderbilt biostatistics clinic for support. The authors also thank Martha W. Tanner, BA, Department of Anesthesiology, Vanderbilt University Medical Center, Nashville, Tennessee, USA, for providing her impressive editorial skills and support as we finalized the manuscript for submission.

DISCLOSURES

Name: Amany K. Alshibli, BS.

Contribution: This author helped with substantial contributions to the conception and design of the work and the analysis and interpretation of data for the work; drafting the work and revising it critically for important intellectual content; and final approval of the version to be published.

Name: Masresha G. Teklehaimanot, BSc.

Contribution: This author helped with substantial contributions to the acquisition and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Rahel Seyoum, MSc.

Contribution: This author helped with substantial contributions to the acquisition and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Gebrehiwot A. Tegu, MSc.

Contribution: This author helped with substantial contributions to the design of the work and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Haftom B. Desta, MSc.

Contribution: This author helped with substantial contributions to the design of the work and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Wu Gong, MS, MD.

Contribution: This author helped with substantial contributions to the analysis and interpretation of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Gosa Tesfaye, MD.

Contribution: This author helped with substantial contributions to the design of the work and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Agenchew N. Tsegaw, MSc.

Contribution: This author helped with substantial contributions to the design of the work and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Abraha Y. Abay, BSc.

Contribution: This author helped with substantial contributions to the design of the work and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Naod B. Etanaa, MSc.

Contribution: This author helped with substantial contributions to the design of the work and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Mulat Mossie, MSc.

Contribution: This author helped with substantial contributions to the design of the work and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Kore M. Benwu, MSc.

Contribution: This author helped with substantial contributions to the design of the work and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Fantahun Tarekegn, MSc.

Contribution: This author helped with substantial contributions to the design of the work and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Hagos G. Gebremedhin, MSc.

Contribution: This author helped with substantial contributions to the design of the work and analysis of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Matthew D. McEvoy, MD.

Contribution: This author helped with substantial contributions to the analysis and interpretation of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Mark W. Newton, MD.

Contribution: This author helped with substantial contributions to the analysis and interpretation of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

Name: Bantayehu Sileshi, MD.

Contribution: This author helped with substantial contributions to the conception and design of the work and the analysis and interpretation of data for the work; revising the work critically for important intellectual content; and final approval of the version to be published.

This manuscript was handled by: Angela Enright, MB, FRCPC.

FOOTNOTES

    REFERENCES

    1. Meara JG, Leather AJ, Hagander L, et al. Global surgery 2030: evidence and solutions for achieving health, welfare, and economic development. Lancet. 2015;386:569–624.
    2. Bainbridge D, Martin J, Arango M, Cheng D; Evidence-based Peri-operative Clinical Outcomes Research (EPiCOR) Group. Perioperative and anaesthetic-related mortality in developed and developing countries: a systematic review and meta-analysis. Lancet. 2012;380:1075–1081.
    3. Hoyler M, Finlayson SR, McClain CD, Meara JG, Hagander L. Shortage of doctors, shortage of data: a review of the global surgery, obstetrics, and anesthesia workforce literature. World J Surg. 2014;38:269–280.
    4. Mazingi D, Navarro S, Bobel MC, Dube A, Mbanje C, Lavy C. Exploring the impact of COVID-19 on progress towards achieving global surgery goals. World J Surg. 2020;44:2451–2457.
    5. Zhu N, Zhang D, Wang W, et al.; China Novel Coronavirus Investigating and Research Team. A novel coronavirus from patients with pneumonia in China, 2019. N Engl J Med. 2020;382:727–733.
    6. Thaddeus S, Maine D. Too far to walk: maternal mortality in context. Soc Sci Med. 1994;38:1091–1110.
    7. COVIDSurg Collaborative C. Elective surgery cancellations due to the COVID-19 pandemic: global predictive modelling to inform surgical recovery plans. Br J Surg. 2020;107:14401449.
    8. Sud A, Jones ME, Broggio J, et al. Collateral damage: the impact on outcomes from cancer surgery of the COVID-19 pandemic. Ann Oncol. 2020;31:1065–1074.
    9. Sud A, Torr B, Jones ME, et al. Effect of delays in the 2-week-wait cancer referral pathway during the COVID-19 pandemic on cancer survival in the UK: a modelling study. Lancet Oncol. 2020;21:1035–1044.
    10. D’Urbano F, Fabbri N, Koleva Radica M, Rossin E, Carcoforo P. Emergency surgery in COVID-19 outbreak: has anything changed? Single center experience. World J Clin Cases. 2020;8:3691–3696.
    11. Tankel J, Keinan A, Blich O, et al. The decreasing incidence of acute appendicitis during COVID-19: a retrospective multi-centre study. World J Surg. 2020;44:2458–2463.
    12. Turanli S, Kiziltan G. Did the COVID-19 pandemic cause a delay in the diagnosis of acute appendicitis? World J Surg. 2021;45:18–22.
    13. Dick L, Green J, Brown J, et al. Changes in emergency general surgery during Covid-19 in Scotland: a prospective cohort study. World J Surg. 2020;44:3590–3594.
    14. De Rosa S, Spaccarotella C, Basso C, et al.; Società Italiana di Cardiologia and the CCU Academy investigators group. Reduction of hospitalizations for myocardial infarction in Italy in the COVID-19 era. Eur Heart J. 2020;41:2083–2088.
    15. Garcia S, Albaghdadi MS, Meraj PM, et al. Reduction in ST-segment elevation cardiac catheterization laboratory activations in the United States during COVID-19 pandemic. J Am Coll Cardiol. 2020;75:2871–2872.
    16. Negreira Caamaño M, Piqueras Flores J, Mateo Gómez C. Impact of COVID-19 pandemic in cardiology admissions. Med Clin (Engl Ed). 2020;155:179–180.
    17. Porreca A, Colicchia M, D’Agostino D, et al. Urology in the time of coronavirus: reduced access to urgent and emergent urological care during the coronavirus disease 2019 outbreak in Italy. Urol Int. 2020;104:631–636.
    18. Sileshi B, Newton MW, Kiptanui J, et al. Monitoring anesthesia care delivery and perioperative mortality in Kenya utilizing a provider-driven novel data collection tool. Anesthesiology. 2017;127:250–271.
    19. Moustakis J, Piperidis AA, Ogunrombi AB. The effect of COVID-19 on essential surgical admissions in South Africa: a retrospective observational analysis of admissions before and during lockdown at a tertiary healthcare complex. S Afr Med J. 2020;110:910–915.
    20. Cano-Valderrama O, Morales X, Ferrigni CJ, et al. Acute care surgery during the COVID-19 pandemic in Spain: changes in volume, causes and complications. A multicentre retrospective cohort study. Int J Surg. 2020;80:157–161.
    21. Biccard BM, Madiba TE, Kluyts HL, et al.; African Surgical Outcomes Study (ASOS) investigators. Perioperative patient outcomes in the African Surgical Outcomes Study: a 7-day prospective observational cohort study. Lancet. 2018;391:1589–1598.
    22. Shigute Z, Mebratie AD, Alemu G, Bedi A. Containing the spread of COVID-19 in Ethiopia. J Glob Health. 2020;10:010369.
    23. Søreide K, Hallet J, Matthews JB, et al. Immediate and long-term impact of the COVID-19 pandemic on delivery of surgical services. Br J Surg. 2020;107:1250–1261.
    24. Kane AD, Paterson J, Pokhrel S, et al. Peri-operative COVID-19 infection in urgent elective surgery during a pandemic surge period: a retrospective observational cohort study. Anaesthesia. 2020;75:1596–1604.
    25. Louie PK, Barber LA, Morse KW, et al. Early peri-operative outcomes were unchanged in patients undergoing spine surgery during the COVID-19 pandemic in New York City. HSS J. 2020;16:18.
    26. Newton MW, Hurt SE, McEvoy MD, et al. Pediatric perioperative mortality in Kenya: a prospective cohort study from 24 hospitals. Anesthesiology. 2020;132:452–460.
    27. Ng-Kamstra JS, Arya S, Greenberg SLM, et al. Perioperative mortality rates in low-income and middle-income countries: a systematic review and meta-analysis. BMJ Glob Health. 2018;3:e000810.
    28. Anderson GA, Ilcisin L, Abesiga L, et al. Surgical volume and postoperative mortality rate at a referral hospital in Western Uganda: measuring the Lancet Commission on Global Surgery indicators in low-resource settings. Surgery. 2017;161:1710–1719.
    29. Rickard JL, Ntakiyiruta G, Chu KM. Associations with perioperative mortality rate at a major referral hospital in Rwanda. World J Surg. 2016;40:784–790.

    APPENDIX

    F5
    Figure A1.:
    Predictor variables and outcomes plotted over time. Mean proportion (for categorical variables) and daily case volume are plotted on the y-axis with day of procedure on the x-axis. LOESS curves with 95% CI are plotted to display the overall trend over time. A, Weekday daily case volume. B, Mean proportion of emergency cases. C, District of origin. D, 28-day perioperative mortality. ACSH indicates Ayder Comprehensive Specialized Hospital; CI, confidence interval; LOESS, locally estimated scatterplot smoothing; TGSH, Tibebe Ghion Specialized Hospital.

    Supplemental Digital Content

    Copyright © 2022 International Anesthesia Research Society