Introduction
Subarachnoid hemorrhages (SAHs) represent 5%–10% of all strokes Reference Rincon, Rossenwasser and Dumont1,Reference Rumalla, Smith, Arnold and Mittal2 and are usually caused by cerebral artery aneurysm rupture. Reference van Gijn, Kerr and Rinkel3,Reference van Lieshout, Dibue-Adjei and Cornelius4 In Canada, the incident admissions for SAH are estimated to be 6.3 per 100 000 person years. For patients deemed salvageable at the time of presentation, the major goal of treatment is to avoid rebleeding by rapidly securing the aneurysm via neurosurgical clipping or angiography. Despite improvements in mortality over the past decade, SAH remains a devastating condition with a case-fatality ratio approaching 36%. Reference Connolly, Rabinstein and Carhuapoma5 In view of this persistent high mortality, a large body of research has sought to identifying risk factors for mortality in SAH.
Age Reference Abulhasan, Alabdulraheem, Simoneau, Angle and Teitelbaum6–Reference Lee, Ouyang and John10 and the severity of neurological injury Reference Hunt and Hess11,12 are well established risk factors for mortality. There is also mounting evidence that pre-existing conditions, including hypertension, cardiac disease, cerebrovascular disease, and renal disease, increase mortality in SAH. Reference Norberg, Odenstedt-Herges, Rydenhag and Oras13,Reference Naval, Kowalski, Chang, Caserta, Carhuapoma and Tamargo14 Non-neurological medical complications occurring during the initial admission also seem to play an important role in SAH mortality. Reference Lantigua, Ortega-Gutierrez and Schmidt9,Reference Wartenberg, Schmidt and Claassen15 For both pre-existing conditions and medical complications however, the extent to which they contribute to mortality and which specific diagnoses matter most remains unclear. In this study, our goal was to assess the effect of pre-existing conditions and medical complications on mortality, at a population level using administrative health data.
Methods
Data Set Source
Administrative health data from the Population Health Record (PopHR) database were used. The PopHR database includes health records for a 25% random sample (open cohort) of the population from the Census Metropolitan Area of Montréal, Canada. Reference Shaban-Nejad, Lavigne, Okhmatovskaia and Buckeridge16 It uses data from the single payer public health insurance, the Régie de l’Assurance Maladie du Québec (RAMQ), which covers over 97% of the Quebec population, 17 along with death certificates and the discharge abstract data from the Maintenance et Exploitation des Données pour l'Étude de la Clientèle Hospitalière (MED-ECHO) database. The information from those three databases was linked by patient deidentified indicators. The study period spanned from 1997 to 2014.
Study Population
Inclusion criteria: Patients were selected based on the documented primary International Classification of Diseases, 9th Revision (ICD-9) diagnosis code of SAH (430); or ICD-10 diagnosis code of SAH (I600-9). Both the MED-ECHO discharge abstract database and RAMQ billings were queried. The Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures (CPP) codes and Canadian Classification of Health Interventions (CCI) codes for surgical clipping and diagnostic and therapeutic angiography were retained to select patients having undergone these procedures (see Appendix 1 under “retained endovascular and neurosurgical procedures” for the full list of procedures and procedure codes). The ICD-9 and CPP codes were used until March 2006, after which the database only includes ICD-10 and CCI codes. We selected in this study only the initial hospitalization episodes (referred to as the index hospitalization) where patients were given the diagnosis of SAH. The initial hospitalization episode included all hospital stays within 3 days of the earliest admission with SAH diagnosis (to include nested admissions, early transfers, and hospital stays where patients may not have been given the diagnosis of SAH), and the subsequent transfers to other institutions. A hospital stay was considered as a transfer, and part of the same hospitalization episode if the patient was admitted within one day of the date of discharge of the previous hospitalization. The positive predictive value (PPV) of using this time interval was found to be 86.4% in a study using administrative health data from the Alberta discharge abstract database. Reference Lichtman, Jones, Leifheit-Limson, Wang and Goldstein18 While the PPV is higher for shorter time intervals (96% for a 6-hour interval), the sensitivity is highest for the larger time interval (98.7%). As the admission and discharge times are not available in our database, and to increase the sensitivity, we opted to use a time interval of less than or equal to one day from the time of discharge Reference Lichtman, Jones, Leifheit-Limson, Wang and Goldstein18 to define our hospitalization episode.
Exclusion criteria: Patients <18 years old were excluded. Patients who were not covered by the RAMQ were not included in this study. Patients not covered by the RAMQ include patients from other provinces covered by their own provincial insurance programs, patients with private insurance visiting Quebec, or patients covered by the Interim Federal Health Program if they are refugees, or from specific groups. 17,19 Patients who did not undergo neurosurgical clipping of an aneurysm or angiography were also excluded. Patients with trauma codes (see Appendix 1, under “trauma codes” for full list of distinct trauma ICD codes used), documented within 14 days of the initial diagnosis of SAH, were excluded, including if the trauma codes were documented during the index hospitalization. We included the cut-off of 14 days prior and after the initial SAH diagnosis because it takes up to 10–14 days for SAH blood to get reabsorbed on imaging, Reference van Gijn and van Dongen20 and the length of stay for traumatic brain injury is usually <14 days. Reference Moore, Stelfox and Turgeon21 See Figure 1 for the flow diagram of the study cohort selection.
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20220630005147256-0718:S0317167121001761:S0317167121001761_fig1.png?pub-status=live)
Figure 1: The Population Health Records database was used to identify 637 adult patients with subarachnoid hemorrhage diagnostic codes, who underwent angiography or neurosurgical clipping of an aneurysm. Patients with concomitant trauma codes were excluded.
Identifying Pre-Existing Conditions
To find pre-existing conditions, the MED-ECHO and RAMQ billing databases were queried for all ICD-9 and ICD-10 codes documented for the patients in our cohort the year prior to the index hospitalization for SAH. Reference Lichtman, Jones, Leifheit-Limson, Wang and Goldstein18 The pre-existing conditions considered for analysis were those deemed severe, as per the Charlson comorbidity index. Reference Norberg, Odenstedt-Herges, Rydenhag and Oras13,Reference Naval, Kowalski, Chang, Caserta, Carhuapoma and Tamargo14 Dyslipidemia and hypertension were added to the list as they were identified as potential risk factors for SAH mortality in previous studies. Reference Lantigua, Ortega-Gutierrez and Schmidt9,Reference Lee, Ouyang and John10,Reference Norberg, Odenstedt-Herges, Rydenhag and Oras13 The full list of pre-existing conditions considered included: dyslipidemia, hypertension, congestive heart failure, myocardial infarction, diabetes, peripheral vascular disease, cerebrovascular disease, rheumatic disease, chronic obstructive pulmonary disease, dementia, peptic ulcer disease, renal disease, any malignancy or metastatic solid tumor, hemiplegia or paraplegia, and liver disease. (see Appendix 1 under “pre-existing medical conditions” for the full list of ICD codes retained).
Identifying Medical Complications
To assess for in-hospital non-neurological medical complications, the MED-ECHO discharge abstract database was queried. All secondary diagnostic codes of the patients recorded during the index hospitalization for SAH were screened, and codes for acute illnesses that were previously identified in the SAH literature as non-neurological medical complications were retained. Reference Rumalla, Smith, Arnold and Mittal2,Reference Lantigua, Ortega-Gutierrez and Schmidt9,Reference Norberg, Odenstedt-Herges, Rydenhag and Oras13,Reference Dasenbrock, Angriman and Smith22–Reference Rivero Rodriguez, Scherle Matamoros, Cue, Miranda Hernandez, Pernas Sanchez and Perez Nellar24 The ICD codes were excluded from the list of retained medical complications if the diagnoses had been recorded in the 12 months prior to the index hospitalization to avoid misclassifying pre-existing conditions as medical complications. This method has been previously published. Reference Lichtman, Jones, Leifheit-Limson, Wang and Goldstein18 The full list of in-hospital non-neurological medical complications retained included: malnutrition, respiratory failure, acute bronchitis, pneumonia, pneumothorax, pulmonary edema, hemothorax, gangrene, coagulation defect, anemia, acute renal failure, urinary tract infection, intestinal obstruction, gastrointestinal hemorrhage, acute liver disease, fluid and electrolyte imbalances, cardiac arrest (successful cardiac resuscitation), heart failure, acute myocardial infarction, and cardiac arrhythmias. (see Appendix 1 under “retained medical complications” for the full list of ICD codes and specific diagnoses).
Method to Establish Disease Severity (Neurological)
The severity of the neurological injury incurred by patients as well as the neurological complications was determined based on the ICD codes for coma, stupor, malaise, fatigue, vertigo, syncope, convulsion, collapse, somnolence, decreased level of consciousness, hydrocephalus, papilledema, cranial nerve abnormalities, paresis/plegia, and aphasia. The procedure codes included to determine the severity of neurological injury were those for ventriculostomy, intracranial pressure monitoring, cerebrospinal fluid diversion, and prolonged intubation. These diagnostic codes and procedure codes have been previously published as indicators of severe neurological injury in SAH Reference Washington, Derdeyn, Dacey, Dhar and Zipfel25 (see Appendix 1 under “severity of neurological injury retained codes” for the full list of ICD codes, and procedure codes retained). In our cohort, if a patient had any of these codes documented, they were grouped in the “severe neurological injury” group and if they had none of the codes queried, they were grouped in the “no severe neurological injury” group.
Statistical Analysis
All statistical analysis was conducted using STATA version 15 statistical software package StataCorp. 2017. Averages, standard deviations, frequencies, and proportions were calculated for age, sex, neurological status, pre-existing conditions, and medical complications (See Table 1). Analysis for the effect of age, sex, severe neurological injury, number of pre-existing conditions, and number of medical complications on the primary outcome was conducted via multivariate binary logistic regression analysis (See Table 1). The primary outcome was mortality at one year, which was determined by using patient death certificates. Age was analyzed as a continuous independent variable, while sex and the severity of neurological injury as categorical independent variables. The pre-existing conditions and medical complications were grouped in four groups, 0, 1, 2, and ≥ 3, and analyzed as categorical independent variables. Similarly, a multivariate binary logistic regression analysis with the binary outcome of having ≥1 in-hospital medical complication was performed, with the independent variables being age, sex, severe neurological injury, and the number of pre-existing conditions (categorical), as seen in Table 2. In Tables 3 and 4, the frequencies and proportions for specific pre-existing conditions and medical complications were calculated and their effects on the primary outcome (one-year mortality) were conducted via univariate binary logistic regression analysis. Finally, if variables were found to be associated with the primary outcome with a p value < 0.05 on the univariate logistic regression analysis in Tables 3 and 4, they were included in a multivariate logistic regression model, which also included, age, sex, and the severity of neurological injury as covariates (See Table 5).
Table 1: Multivariate logistic regression for one-year mortality
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20220630005147256-0718:S0317167121001761:S0317167121001761_tab1.png?pub-status=live)
CI = confidence interval; n = number = OR odds ratio; SD = standard deviation.
Hosmer–Lemeshow goodness of fit: 0.95.
Table 2: Multivariate logistic regression for ≥1 medical complications during hospitalization
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20220630005147256-0718:S0317167121001761:S0317167121001761_tab2.png?pub-status=live)
CI = confidence interval; OR = odds ratio.
Hosmer–Lemeshow goodness of fit: 0.26.
Table 3: Frequency table for pre-existing conditions and univariate regression for mortality at one year, grouped by pre-existing condition
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20220630005147256-0718:S0317167121001761:S0317167121001761_tab3.png?pub-status=live)
CI = confidence interval; OR = odds ratio.
* No deaths occurred in this group, so univariate logistic regression for one-year mortality not performed.
Table 4: Frequency table for medical complications and univariate regression for mortality at one year, grouped by medical complication
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20220630005147256-0718:S0317167121001761:S0317167121001761_tab4.png?pub-status=live)
CI = confidence interval; OR = odds ratio.
* No deaths occurred in this group, so univariate logistic regression for one-year mortality not performed.
Table 5: Multivariate logistic regression for mortality at one year for specific pre-existing conditions and medical complications
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20220630005147256-0718:S0317167121001761:S0317167121001761_tab5.png?pub-status=live)
CI = confidence interval; OR = odds ratio.
Hosmer–Lemeshow goodness of fit: 0.41.
Results
Descriptive Statistics: Demographics, Hospital Stay, and Discharge Status
The total number of patients in this cohort was 637. The youngest patient being 18 years old, and the oldest patient being 92 years old. The average length of stay of the hospitalization episode was 41.07 days (standard deviation 50.8 days), ranging from a minimum of 0 days to a maximum of 396 days. Fifty-one patients (8.0%) died at the institution where they were initially admitted, and 34 (5.3%) died over the course of the remainder of the hospitalization episode after transfer to another acute care facility. In terms of discharge status at the end of the hospitalization episode, 73.62% of patients were discharged home or home with services, 8.63% were discharged to a long-term care facility, 2.5% were transferred to another acute care hospital, 0.78% were discharged to a rehabilitation center, 0.47% left against medical advice, and 0.64% were discharged to another institution which was not specified.
Mortality: Effect of the Number of Pre-Existing Conditions and Medical Complications
Over the course of the whole hospitalization episode, 85 patients died, bringing the overall in-hospital mortality to 13.3%. At one year, a total of 94 patients had died, for an overall one-year mortality rate of 14.76%.
As seen in Table 1, the average age was 54.63 years (standard deviation 13.71 years). The majority of the patients were female (n = 409, 64.21%). Neither increasing age nor male sex was associated with increased mortality at one year. Severe neurological injury was associated with a 3.42-fold increase in mortality [95% CI 1.90–6.13]. Two hundred and twenty-four patients (35.16%) had at least one pre-existing conditions. While having 1 or 2 pre-existing conditions did not significantly increase mortality at 1 year, having 3 or more pre-existing conditions was associated with increased one-year mortality OR 3.74, [95% CI 1.25–9.57]. Two hundred and eighty-four patients (44.58%) had at least one medical complication over the course of their hospitalization episode. While having one medical complication was not associated with increased mortality, having 2 or ≥3 complications was associated with increased one-year mortality, OR 2.42 [95% CI 1.25–4.69] and OR 2.69 [95% CI 1.43–5.07], respectively. As seen in Table 2, in a multivariate logistic regression model controlling for age, sex, and the severity of neurological injury, having 1, 2, and ≥3 pre-existing conditions was associated with a 1.51-, 2.03-, and 3.78-fold increase in odds of patients having experienced at least 1 medical complication, respectively.
Mortality: Effect of Specific Pre-Existing Conditions and Medical Complications
As described in Table 3, malignancy (5.18%), hypertension (16.80%), diabetes (7.69%), and chronic pulmonary disease (6.44%) were found to be the most common pre-existing conditions and malignancy, diabetes, myocardial infarction, congestive heart failure, renal disease, and cerebrovascular disease were associated with increased one-year mortality on univariate logistic regression. As seen in Table 4, pneumonia (13.50%), fluid and electrolyte imbalances (12.87%), cardiac arrhythmias (7.38%), urinary tract infections (14.76%), and anemia (8.63%) were the most common medical complications recorded and sepsis, shock, respiratory failure, pneumonia, acute renal failure, fluid and electrolyte imbalance, acute myocardial infarction and cardiac arrhythmias were associated with increased one-year mortality on univariate logistic regression.
As demonstrated in Table 5, on multivariate logistic regression, controlling for age, sex, and the severity of the neurological injury, while none of the pre-existing conditions were found to be associated with increased mortality, sepsis, respiratory failure, and cardiac arrhythmias were found to be associated with increased mortality at one year.
Discussion
In this study, we have measured the effect of pre-existing conditions and medical complications on SAH mortality, while controlling for age, sex, and the severity of neurological injury. Of the 637 patients in our cohort, 94 died within a year of their initial SAH diagnosis (14.76%), of which 85 patients died during the hospitalization episode (13.3%). These mortality rates are similar to those previously reported for in-hospital and one-year mortalities (12.6%–13.9%), for SAH patients who were admitted to the hospital and underwent angiography or neurosurgical clipping of an aneurysm. Reference Stienen, Germans and Burkhardt26,Reference Ronkainen, Niskanen, Rinne, Koivisto, Hernesniemi and Vapalahti27 Male sex was not significantly associated with mortality in our cohort OR, 1.45 [95% CI 0.90–2.33]. This finding is consistent with recent studies which show sex does not appear associated with mortality, when controlling for other risk factors including the severity of the neurological injury and age. Reference Abulhasan, Alabdulraheem, Simoneau, Angle and Teitelbaum6–Reference Lantigua, Ortega-Gutierrez and Schmidt9,Reference Claassen, Vu and Kreiter28
In our cohort, age was not found to be associated with mortality OR, 1.00 [95% CI 0.99–1.02]. While several studies looking at new SAH patients have found that increasing age is associated with increased mortality, Reference Abulhasan, Alabdulraheem, Simoneau, Angle and Teitelbaum6–Reference Lee, Ouyang and John10 a large institutional study looking exclusively at patients who were treated with angiography or neurosurgical clipping found that increasing age was not associated with increased mortality. Reference Degos, Gourraud and Tursis29 Our findings are consistent with this study and may be due to patients selected to undergo aneurysmal treatments are deemed medically and neurologically salvageable and may share similar clinical characteristics.
Compared to patients with no pre-existing conditions, having 1 or 2 pre-existing conditions did not significantly increase mortality at 1 year but having ≥3 pre-existing conditions was associated with increased one-year mortality OR 3.74, [95% CI 1.25–9.57]. Compared to patients with no medical complications, having 2 or ≥3 complications was associated with increased one-year mortality, OR 2.42 [95% CI 1.25–4.69], and OR 2.69 [95% CI 1.43–5.07], respectively, after controlling for age, sex, severity of neurological injury, and the number of pre-existing conditions.
When looking at which specific pre-existing conditions affect mortality, Norberg et al., reported patients with a history of cardiac disease HR, 2.31 [95% CI 1.26–4.22], cerebrovascular disease HR, 2.35 [95% CI 1.19–4.66], and renal disease HR, 3.23 [95% CI 1.18–8.96], showed higher risks of mortality [36]. In our cohort, we also reported cardiac disease, including congestive heart failure OR, 3.98 [95% CI 1.10–14.37], and myocardial infarction OR, 4.78 [95% CI 1.26–18.15], cerebrovascular disease OR, 2.99 [95% CI 1.00–8.97], and renal disease OR, 3.98 [95% CI 1.10–14.37, to be associated with increased mortality on univariate logistic regression analysis (Table 3). They had however also reported hypertension to be associated with increased mortality, HR, 1.74 [95% CI 1.18–2.57], which we did not, and they had not reported malignancy or diabetes to be associated, which we did OR, 2.29 [95% CI 1.03–5.09] and OR, 2.26 [95% CI 1.15–4.44], respectively. These differences may be due to several methodological differences including the use of chart review in the Norberg study to determine pre-existing conditions, a different geographic location (Montréal versus Sweden), and the fact we selected only patients who had undergone angiography or neurosurgical treatment of the aneurysm. Previous research on SAH has shown that geographic location can affect the risk of rupture of aneurysms. This is particularly true for Scandinavian countries which have been found to have a high risk of aneurysmal rupture when compared to North American countries. Reference Bijlenga, Gondar and Schilling30 Of note, on multivariate logistic regression, controlling for age, sex, the severity of neurological injury and medical complications (Table 5), none of the pre-existing conditions was found to be associated with increased mortality. As seen in Table 2, pre-existing conditions are associated with increased odds of patients having medical complications.
As seen in Table 5, we completed a multivariate logistic regression, controlling for age, sex, the severity of neurological injury, and the pre-existing conditions and medical complications that were found to be associated with increased mortality on univariate logistic regression (p < 0.05), in Tables 3 and 4. Looking at specific medical complications, respiratory failure was associated with increased mortality in our cohort OR, 4.13 [95% CI 1.33–12.80]. This finding is consistent with Solenski et al., who reported that pulmonary complications were the most common cause of death in their cohort. Reference Solenski, Haley and Kassell31 Sepsis was also associated with increased mortality OR, 4.95 [95% CI 1.37–17.83]. This is consistent with previous groups that had reported hypotension and fever to be associated with mortality. Reference Lantigua, Ortega-Gutierrez and Schmidt9,Reference Wartenberg, Schmidt and Claassen15 Cardiac arrythmias were also associated with increased mortality OR, 2.86 [95% CI 1.39–5.89]. This is consistent with a large prospective study, which reported cardiac arrythmias to be associated with increased mortality, even when controlling for age, and the severity of neurological injury. Reference Wartenberg, Schmidt and Claassen15 This is of particular interest because previous authors had reported that while common, cardiac arrythmias rarely lead to deaths. Reference Solenski, Haley and Kassell31 Shock, pneumonia, acute renal failure, fluid and electrolyte imbalance, acute myocardial infarction were not found to be associated with increase mortality, on multivariate logistic regression analysis, which could be due to the relatively low power of the study. Fever, hypotension, hyperglycemia, and anemia were reported to be associated with increased mortality by previous authors. Reference Lantigua, Ortega-Gutierrez and Schmidt9,Reference Wartenberg, Schmidt and Claassen15 In our cohort, these diagnoses were not associated with increased mortality. This could be due to the fact that these diagnoses are largely based on changes in vital signs and specific laboratory findings which may be more appropriate to collect in institutional cohorts, where these data can be obtained directly.
In our study, we found having 1, 2, and ≥3 pre-existing conditions was associated with a 1.51-, 2.03-, and 3.78-fold increase in odds of patients having experienced at least 1 medical complication, respectively (Table 2). This finding suggests that pre-existing conditions may increase mortality in SAH patients by increasing medical complications. While previous studies had described that pre-existing conditions were associated with increased mortality, the mechanism by which they do so is not well understood. Our study shows that it may be via an increase in medical complications. Further research should aim to better characterize the association between specific pre-existing conditions and specific medical complications.
This study has limitations. The identification of the diagnoses and of the procedures patients underwent was based on ICD-9, ICD-10, CPP, and CCI codes, with their associated risks of misclassification. This is a known limitation of administrative health data. That said, the codes for SAH have been validated by previous research and found to have very good accuracy. Reference Rumalla, Smith, Arnold and Mittal2 We are therefore quite confident our patients were true aneurysmal SAH patients, especially considering we excluded patients with trauma codes and included only patients who had undergone angiography or neurosurgical clipping of an aneurysm. To avoid misclassification of pre-existing conditions, we have strictly used the same diagnostic codes previous authors had used when identifying pre-existing conditions using administrative health data. Reference Deyo, Cherkin and Ciol32,Reference Quan, Sundararajan and Halfon33
As with all administrative databases, there are limitations in the collection of clinical information. Specifically for our project, we used administrative health data to determine acute medical events and complications that occurred during the hospitalization of patients newly diagnosed with SAH. One of the risks of doing so is misclassifying concomitant ICD codes as acute when they are in fact pre-existing conditions. We took several steps to mitigate this risk. First, we excluded from the ICD codes for medical complications all the codes that had been previously documented as pre-existing conditions in our cohort. Second, we examined every diagnostic codes and only included those that were coding for diagnoses that had been classified as acute medical complications in the SAH literature. Reference Lantigua, Ortega-Gutierrez and Schmidt9,Reference Wartenberg, Schmidt and Claassen15,Reference Solenski, Haley and Kassell31 Third, we have included in appendix 1 all the diagnostic codes that have been retained as medical complications. While there is a risk of misclassification, this methodology has permitted us to analyze the impact of important acute medical events on mortality.
Similarly, while the gold standard to assess the severity of the neurological injury is the use of the HH and WFNS scores, the variables used in our cohort to assess the extent of the neurological injury and neurological complications have been previously published. Reference Washington, Derdeyn, Dacey, Dhar and Zipfel25 Furthermore, the odds of mortality between the severe neurological injury group and the non-severe neurological group in our cohort OR, 3.42 [95% CI 1.90–6.13], mirror those published in institutional studies comparing HH 1–3 to HH 4 groups OR, 4.08 [95% CI 1.55–7.30]. Reference Lee, Ouyang and John10 One other limitation is the risk of misclassifying pre-existing conditions as medical complications. For medical complications occurring in hospital, to mitigate this risk, only codes that had not been given to patients the year prior to admission were retained and only a priori variables of interest based on the SAH literature were selected. Reference Rumalla, Smith, Arnold and Mittal2,Reference Lantigua, Ortega-Gutierrez and Schmidt9,Reference Norberg, Odenstedt-Herges, Rydenhag and Oras13,Reference Dasenbrock, Angriman and Smith22–Reference Rivero Rodriguez, Scherle Matamoros, Cue, Miranda Hernandez, Pernas Sanchez and Perez Nellar24 Another risk when assessing in hospital medical complications, interventions, and mortality for hospitalized patients using administrative health data is censoring due to patients transfers. To mitigate this risk, we constructed hospitalization episodes which included all transfers within 3 days of the initial SAH admission, as well as the subsequent transfer to another acute care hospital as part of the same hospitalization episode.
One limitation of the statistical analysis is that the models presented in Tables 1, 2 and 5 were not corrected for multiple testing. However, the variable selection described in the Methods section is commonly used, and correcting for multiple testing may be associated with an increased risk of false negatives.
Through the findings in this study, we have shown that pre-existing conditions have an important impact on SAH mortality. The effect measures provided can help physicians be better prepared when dealing with SAH patients with multiple comorbidities, especially those with malignancies, diabetes, congestive heart failure, renal disease, and cerebrovascular disease. By knowing the effect of these pre-existing conditions, on mortality and the increased risk of developing medical complications, physicians may opt for closer observation, monitoring, and laboratory testing for these patients. Importantly, these patients appear at risk of in-hospital medical complications, namely sepsis, respiratory failure, and cardiac arrythmias. Increased monitoring and aggressive management is warranted. Furthermore, public health measures to reduce pre-existing conditions should continue to be put forth. Last, this knowledge on the effect of medical complications on mortality can help physicians guide patients and families as they occur during the initial hospitalization.
One of the major strengths of this study is the generalizability of our findings. The data collected by the PopHR database spanned 17 years, and multiple institutions. Furthermore, the known risk factors for SAH mortality were controlled for via binary multivariate logistic regression models, and the exposures included were collected over the course of the year preceding the initial hospitalization as well as the whole initial hospitalization episode. We opted to include in our analysis only patients who had undergone angiography or neurosurgical clipping of their aneurysm. We did so in order to establish an effect measure for pre-existing conditions and medical complications for patients that were deemed salvageable from a neurological standpoint at the time of admission. While this limits the generalizability of our findings to only patients undergoing procedures, it provides clinically relevant information for physicians caring for this patient population. Taken together, these findings can be generalized to patients newly diagnosed with SAH, admitted alive to an acute care hospital, and who undergo aneurysmal clipping or angiography, in a North American hospital, for patients covered by health insurance.
Conclusions
Pre-existing conditions and medical complications have been found to be associated with increased one-year mortality in SAH patients undergoing aneurysmal clipping or angiography. In particular, in-hospital complications are sepsis, respiratory failure, and cardiac arrythmias. Pre-existing conditions are associated with increased medical complications.
Acknowledgements
The authors would also like to acknowledge the contributions of Guido Powell and Dr. Aman Verma. We would also like to acknowledge the McGill neurosurgery department for accommodating time for this research during residency. The authors would like to acknowledge funding from the Canadian Institute of Health Research (CIHR), Frederick Banting and Charles Best Canada Graduate Scholarships – Master’s (CGS M) award recipient, in support of this project.
Statement of Authorship
SS identified a gap in knowledge, completed the literature review, and proposed the research questions. DB and DP helped narrow the scope of the research questions, along with setting clear objectives. DB has been instrumental in providing access to the MED-ECHO database, the Régie de l’Assurance Maladie du Québec (RAMQ) billing codes database, and death certificate database. DB and DP have helped supervise and revised the writing of the thesis and manuscript. DB and DP have helped select the appropriate statistical tools for data analysis. SD has provided guidance regarding the clinical relevance of the thesis. The research methods were written by SS, with editorial assistance from DB and DP. The data analysis and interpretation were completed by SS.
Disclosures
The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/cjn.2021.176