Mortality and Time to Endoscopy in Upper GI Hemorrhage
Mortality and Time to Endoscopy in Upper GI Hemorrhage
We analysed demographic risk factors for upper GI haemorrhage in the largest and most diverse patient database in the US. Previous studies had focused only on geographical or health system – based populations, or just one type of UGIH – either AVH or NVUGIH. We concluded that in-patient mortality for AVH and NVUGIH were 5.1% and 11.1% respectively. In-patient mortality due to AVH and NVUGIH increased with age, comorbidity, male gender, and for those who do not receive an OGD within 1 day of hospital admission. African American race predicted increased in-patient mortality due to AVH, but it was not a significant risk factor in mortality due to NVUGIH.
The timing of endoscopic intervention has received much attention in recent years. Tsoi et al evaluated the optimal timing of early OGD and concluded that no evidence exists to perform early endoscopy within 12 h of presentation. However, endoscopy within 24 h of presentation had clinical benefits in risk assessment, reducing length of stay, reducing surgical risk, and preventing in-patient mortality. The American Society for Gastrointestinal Endoscopy stated that 'early endoscopy (within 24 h of hospital admission) had a greater impact than delayed endoscopy on length of hospital stay and requirements for blood transfusion'. Based on the results for both AVH and NVUGIH, the populations at greatest risk of not receiving endoscopy within 1 day of admission were African American women, Medicaid patients, those >80 years old and patients with more than three comorbidities. If patients did not receive an OGD within 1 day of admission, mortality increased from 8.25% to 15.3% for AVH and 2.5% to 6.6% for NVUGIH. Patients with private insurance had less in-patient mortality and were more likely to have received OGD within 1 day of admission for AVH and NVUGIH compared to patients with Medicare and Medicaid. Because urgent endoscopy improves outcomes for patients at risk for re-bleeding and death, we concluded that endoscopy within 1 day is a standard of care – a significant difference observed in our study population.
Our analysis found that African Americans had increased mortality risk from AVH compared to White Americans. In addition, African Americans were less likely to receive OGD within 1 day of admission for both AVH and NVUGIH. Multiple corrections were proposed to account for this difference, and we carried out analyses to attempt to account for the excess mortality and delayed endoscopy. When we corrected for primary payer, comorbidity, hospital size, hospital ownership and OGD within 1 day (for mortality), these differences persisted. Furthermore, when we corrected for time to endoscopy, increased mortality still persists among African Americans, suggesting that time to endoscopy is not the primary cause of increased mortality in this group.
There is much debate about overall healthcare outcomes between African Americans and White Americans. Smedley et al found a difference in the quality of health care between white and minority populations. Compared to non-Hispanic whites, African Americans have higher rates of morbidity and mortality for nearly all conditions, and they are increasingly less likely to have health insurance and receive job-based health coverage. Pippins et al highlighted a racial disparity where African Americans had higher mortality than White Americans in UGIH. The authors suggested that hospitals serving larger populations of African Americans provided poorer quality of care for both African American and White patients.
Numerous theories have been proposed to explain the observed racial discrepancies, such as: (i) less access to preventative and primary care services; (ii) a higher rate of system-specific symptoms or complications in non white patients; (iii) socioeconomic differences; (iv) less insurance coverage; (v) language barriers; (vi) genetic bias; and (vii) behavioural patterns. Studies performed by Volpp and Polsky, found that race was an independent risk factor of unadjusted 30-day and 2-year mortality rates for six high-severity conditions in the VA system, with African Americans having lower 30-day mortality rates but higher 2-year mortality rates. Gordon et al. argued that more equal access to care (such as integrated systems like the VA) may eliminate racial disparities. However, there are limitations to using the VA system. There is less disparity among this population in terms of socioeconomic status, lifestyle and healthcare costs compared to the general population, which utilises private insurance, Medicaid and Medicare. Analysing racial differences across large, diverse populations, as they pertain to selected medical conditions such as UGIH, remains a challenge to social researchers.
Several limitations were encountered in this study. One crucial factor in analysing early endoscopic intervention is to determine the hemodynamic stability of the patients. It is not possible to reliably extrapolate data in this detail from the NIS database, so all patients were included. Procedure-specific details and operator experience were not stored in the NIS database. Since the NIS is based on discharge data, we were unable to determine patient specific follow-up after discharge or report on patients that were only in observational status in the Emergency Department. In addition, the dataset includes all patients who were discharged with an UGIH; thus, there is no way to differentiate hospital-acquired vs. UGIH that resulted in admission to the hospital. For patients with UGIH that occurred during hospitalisation, there is no way to accurately determine time to endoscopy. Procedure-specific details are not included in the NIS database. Therefore, findings that were predictive of future re-bleeding (active squirting blood, clean-based ulcers, adherent clot, etc.) were not incorporated in our modelling. Bleeding episodes are often self-limited, which may leave the final coding assignment subject to interpretation of the coders/managers, and may be incorrectly coded. Due to the wide range of ICD-9 diagnosis codes for upper GI haemorrhage, it is possible that we included patients who were not a part of the target population. Moreover, patients that have complicated hospital courses, critically ill, or have contraindications to urgent endoscopy have not been sub-categorised and remain part of model. Also, specific details of endoscopic treatment and pharmacological treatments are not available in the NIS database and therefore could not be analysed. Finally, the time to endoscopy was coded in days in the NIS database, rather than hours. Thus, it is possible that an OGD performed within a day or less of admission could reflect a length of time longer than 24 h from initial presentation or onset of symptoms.
We conclude that there is an increased mortality risk in African Americans compared to other racial groups with AVH. Racial differences also existed in receipt of OGD within the first day of hospitalisation for both AVH and NVUGIH. To date, no study has utilised such a large, diverse database to describe these significant demographic risk factors in UGIH outcomes. African American race should be viewed as a risk factor for higher mortality in AVH and African Americans are less likely to receive an OGD within 1 day of admission for any UGIH.
Discussion
We analysed demographic risk factors for upper GI haemorrhage in the largest and most diverse patient database in the US. Previous studies had focused only on geographical or health system – based populations, or just one type of UGIH – either AVH or NVUGIH. We concluded that in-patient mortality for AVH and NVUGIH were 5.1% and 11.1% respectively. In-patient mortality due to AVH and NVUGIH increased with age, comorbidity, male gender, and for those who do not receive an OGD within 1 day of hospital admission. African American race predicted increased in-patient mortality due to AVH, but it was not a significant risk factor in mortality due to NVUGIH.
The timing of endoscopic intervention has received much attention in recent years. Tsoi et al evaluated the optimal timing of early OGD and concluded that no evidence exists to perform early endoscopy within 12 h of presentation. However, endoscopy within 24 h of presentation had clinical benefits in risk assessment, reducing length of stay, reducing surgical risk, and preventing in-patient mortality. The American Society for Gastrointestinal Endoscopy stated that 'early endoscopy (within 24 h of hospital admission) had a greater impact than delayed endoscopy on length of hospital stay and requirements for blood transfusion'. Based on the results for both AVH and NVUGIH, the populations at greatest risk of not receiving endoscopy within 1 day of admission were African American women, Medicaid patients, those >80 years old and patients with more than three comorbidities. If patients did not receive an OGD within 1 day of admission, mortality increased from 8.25% to 15.3% for AVH and 2.5% to 6.6% for NVUGIH. Patients with private insurance had less in-patient mortality and were more likely to have received OGD within 1 day of admission for AVH and NVUGIH compared to patients with Medicare and Medicaid. Because urgent endoscopy improves outcomes for patients at risk for re-bleeding and death, we concluded that endoscopy within 1 day is a standard of care – a significant difference observed in our study population.
Our analysis found that African Americans had increased mortality risk from AVH compared to White Americans. In addition, African Americans were less likely to receive OGD within 1 day of admission for both AVH and NVUGIH. Multiple corrections were proposed to account for this difference, and we carried out analyses to attempt to account for the excess mortality and delayed endoscopy. When we corrected for primary payer, comorbidity, hospital size, hospital ownership and OGD within 1 day (for mortality), these differences persisted. Furthermore, when we corrected for time to endoscopy, increased mortality still persists among African Americans, suggesting that time to endoscopy is not the primary cause of increased mortality in this group.
There is much debate about overall healthcare outcomes between African Americans and White Americans. Smedley et al found a difference in the quality of health care between white and minority populations. Compared to non-Hispanic whites, African Americans have higher rates of morbidity and mortality for nearly all conditions, and they are increasingly less likely to have health insurance and receive job-based health coverage. Pippins et al highlighted a racial disparity where African Americans had higher mortality than White Americans in UGIH. The authors suggested that hospitals serving larger populations of African Americans provided poorer quality of care for both African American and White patients.
Numerous theories have been proposed to explain the observed racial discrepancies, such as: (i) less access to preventative and primary care services; (ii) a higher rate of system-specific symptoms or complications in non white patients; (iii) socioeconomic differences; (iv) less insurance coverage; (v) language barriers; (vi) genetic bias; and (vii) behavioural patterns. Studies performed by Volpp and Polsky, found that race was an independent risk factor of unadjusted 30-day and 2-year mortality rates for six high-severity conditions in the VA system, with African Americans having lower 30-day mortality rates but higher 2-year mortality rates. Gordon et al. argued that more equal access to care (such as integrated systems like the VA) may eliminate racial disparities. However, there are limitations to using the VA system. There is less disparity among this population in terms of socioeconomic status, lifestyle and healthcare costs compared to the general population, which utilises private insurance, Medicaid and Medicare. Analysing racial differences across large, diverse populations, as they pertain to selected medical conditions such as UGIH, remains a challenge to social researchers.
Several limitations were encountered in this study. One crucial factor in analysing early endoscopic intervention is to determine the hemodynamic stability of the patients. It is not possible to reliably extrapolate data in this detail from the NIS database, so all patients were included. Procedure-specific details and operator experience were not stored in the NIS database. Since the NIS is based on discharge data, we were unable to determine patient specific follow-up after discharge or report on patients that were only in observational status in the Emergency Department. In addition, the dataset includes all patients who were discharged with an UGIH; thus, there is no way to differentiate hospital-acquired vs. UGIH that resulted in admission to the hospital. For patients with UGIH that occurred during hospitalisation, there is no way to accurately determine time to endoscopy. Procedure-specific details are not included in the NIS database. Therefore, findings that were predictive of future re-bleeding (active squirting blood, clean-based ulcers, adherent clot, etc.) were not incorporated in our modelling. Bleeding episodes are often self-limited, which may leave the final coding assignment subject to interpretation of the coders/managers, and may be incorrectly coded. Due to the wide range of ICD-9 diagnosis codes for upper GI haemorrhage, it is possible that we included patients who were not a part of the target population. Moreover, patients that have complicated hospital courses, critically ill, or have contraindications to urgent endoscopy have not been sub-categorised and remain part of model. Also, specific details of endoscopic treatment and pharmacological treatments are not available in the NIS database and therefore could not be analysed. Finally, the time to endoscopy was coded in days in the NIS database, rather than hours. Thus, it is possible that an OGD performed within a day or less of admission could reflect a length of time longer than 24 h from initial presentation or onset of symptoms.
We conclude that there is an increased mortality risk in African Americans compared to other racial groups with AVH. Racial differences also existed in receipt of OGD within the first day of hospitalisation for both AVH and NVUGIH. To date, no study has utilised such a large, diverse database to describe these significant demographic risk factors in UGIH outcomes. African American race should be viewed as a risk factor for higher mortality in AVH and African Americans are less likely to receive an OGD within 1 day of admission for any UGIH.