Introduction
Cerebral venous thrombosis (CVT) refers to thrombosis of the intracranial venous sinuses, deep cerebral veins, or cortical veins. CVT diagnosis can often be delayed with previous studies reporting an average time to diagnosis of 4–9 days from symptom onset. Reference Liberman, Gialdini, Bakradze, Chatterjee, Kamel and Merkler1–Reference Timoteo, Inacio, Machado, Pinto and Parreira3 There are multiple reasons that contribute to diagnostic delays. First, CVT is a rare condition, with an incidence of 10–20 per million Reference Zhou, Yu, Ngo, Hill and Field4 and thus may not be front-of-mind on a differential diagnosis. Second, presentations of CVT may be different from other stroke types, sometimes presenting with isolated increased intracranial pressure symptoms (headache, visual changes, nausea and vomiting) or seizures without additional sudden-onset focal features. Reference Ferro, Lopes, Rosas and Fontes2,Reference Ferro, Canhão, Stam, Bousser and Barinagarrementeria5 Finally, CVT most commonly affects younger women, and both women and young adults most commonly face delays in diagnosis of stroke. Reference Newman-Toker, Moy, Valente, Coffey and Hines6,Reference Eddelien, Butt, Christensen, Danielsen and Kruuse7
A US study from 2005 to 2013 showed that 3.6% of patients with CVT were first assessed for headache or seizure within the emergency department prior to a subsequent visit where CVT was diagnosed. Those patients with a delay in diagnosis of CVT had a longer length of stay in hospital. Reference Liberman, Gialdini, Bakradze, Chatterjee, Kamel and Merkler1 However, within the Canadian context, the specific issue of missed opportunities for earlier diagnosis has not been well documented. Examining Canadian data may clarify the potential scope of the issue to be addressed and can be the first step in developing educational strategies and clinical pathways to improve the timeliness of CVT diagnosis. We used administrative healthcare data from British Columbia (BC) to create a population-based inclusive cohort of adult CVT patients. We then examined encounters with physicians in the 3 and 7 days prior to hospital admission for those with a new diagnosis of CVT between 2000 and 2017.
Methods
Study design, setting and data sources
We conducted a population-based study using linked patient-level administrative and registry data for adults ≥18 years hospitalized with CVT in British Columbia (BC) between 2000 and 2017. Data linkages and secure data storage were conducted by Population Data BC, a multi-university, data and education resource that helps facilitate population health research by providing timely access to de-identified patient-level data in BC. 8 Access to data provided by the Data Stewards is subject to approval but can be requested for research projects through the Data Stewards or their designated service providers. The following data sets were used in this study: hospital separations Discharge Abstract Database, Medical Services Plan (MSP), MSP Consolidation File, BC Cancer Registry, and BC Perinatal Data Registry. You can find further information regarding these data sets by visiting the PopData project webpage at: https://my.popdata.bc.ca/project_listings/19-118/collection_approval_dates. All inferences, opinions, and conclusions drawn in this publication are those of the authors, and do not reflect the opinions or policies of the Data Stewards. The study was approved by the Clinical Research Ethics Board at the University of British Columbia (H18-03598).
Study population and analytic dataset/variables
The study cohort was created by Population Data BC by scanning the Discharge Abstract Database 9 for hospital admissions that contained an ICD-9 (325.0, 437.6, 671.5) or ICD-10-CA (G08, I63.6, I67.6, O22.5, O87.3) diagnosis code for CVT which has previously been validated for BC-based data. Reference Zhou, Yu, Hall, Hill and Field10 Inpatient admissions contiguous in time to this initial CVT admission (i.e., on the same day or the day before or after) were concatenated to capture the individual’s entire index admission (e.g., including patients admitted to one hospital but transferred to another). The date of CVT diagnosis was defined as the admission date for the first hospital admission. Demographic information (age, sex) was collected from the MSP Consolidation File. 11 Detailed methodology for developing the cohort, alongside description of baseline characteristics, have been published previously. Reference Zhou, Yu, Ngo, Hill and Field4
Cohort members’ cancer status was determined using data from the BC cancer registry dataset. Cancer-related CVTs were defined as a CVT diagnosis within 12 months of registration into the registry (squamous and basal cell skin carcinomas were excluded). 12 Peripartum status was determined using registration within the BC perinatal data registry, which contains data from an estimated 99% of all live births and stillbirths at ≥20 weeks or with a birthweight of ≥500 g. Peripartum CVT was defined as a CVT diagnosis within 3 months of delivery. 13 Peripartum, trauma and cancer status were supplemented from manual review of the first five ICD diagnosis codes from the patient’s index CVT admission.
Pre-hospitalization health care utilization
Using the MSP physician billing dataset, 14 we examined the number of each individual’s unique encounters with a health care provider within the 3 and 7 days prior to index hospital admission for CVT. We selected this timeframe because of previous publications on delays to diagnosis. Reference Liberman, Gialdini, Bakradze, Chatterjee, Kamel and Merkler1–Reference Timoteo, Inacio, Machado, Pinto and Parreira3 Unique encounters were identified by removing multiple billed services provided on the same date from physicians with the same specialty code. Diagnostic service entries such as radiology, pathology, medical microbiology and nuclear medicine services were excluded. Services provided on the day of CVT hospitalization were excluded as these were assumed to be related to the index admission. A sensitivity analysis was conducted by further removing services from the day prior to admission to remove encounters that may have directly resulted in the index admission (i.e., assessment by a physician and referral for admission within a single encounter occurring over two calendar days due to wait times past midnight). MSP billings were then divided into inpatient and outpatient encounters. Inpatient encounters were visits on a date where a patient was admitted at any hospital then discharged without a diagnosis of CVT. If a patient was not admitted to hospital on the date of the encounter, the encounters were classified as outpatient encounters.
Statistical analysis
Associations between patient factors (age, sex, and CVT risk factors such as peripartum status, use of estrogen, malignancy status, traumatic cause of CVT) and pre-CVT health care utilization were examined first through bivariate analysis through Chi-squared and t-tests where appropriate, and then through multivariable logistic regression modeling while controlling for covariates (age, sex, time epoch of admission, and CVT risk factors). Age was binned into categories for multivariable analysis due to evidence of non-linearity. Using encounter frequency, we also summarized the most common physician specialties involved in outpatient and inpatient encounters within 7 days of admission.
Data in small subgroups (i.e., non-zero low numbers with n < 5) were suppressed to avoid the identification of specific patients in compliance with the data use agreement with Population Data BC. All statistical analyses were performed using RStudio version 4.0.5.
Results
Study cohort
A total of 554 individuals were diagnosed with CVT between 2000 and 2017 (mean age 50.9 years, SD 19.4; 55.4% female). Baseline characteristics of this cohort have been described previously, Reference Zhou, Yu, Ngo, Hill and Field4 and are presented in Table 1 for reference.
Table 1. Cohort characteristics (n = 554)
* Including all non-traumatic intracerebral hemorrhagic (I60-I62).
CVT = cerebral venous thrombosis; IQR = interquartile range; SD = standard deviation.
Overall and stratified by presence of at least one outpatient encounter within the 7 or 3 days prior to their CVT diagnosis. Values in parentheses represent row percentages.
With respect to outpatient encounters (Figure 1, Panel A), 57.9% had ≥1 encounter within 7 days prior to CVT admission (median = 1 day, IQR = 2) and 46.8% had ≥1 outpatient encounter within 3 days prior to CVT admission (median = 0 days, IQR = 1). The timeframe for these encounters is shown in Figure 1, Panels C and D. Encounters from the day prior to CVT admission only accounted for 32.1% of all outpatient encounters within 7 days.
Figure 1. Frequency and timing of healthcare encounters prior to cerebral venous thrombosis (CVT) admission. Legend: Panel (A) Percentage of patients with or without outpatient encounters within the 7 and 3 days prior to their CVT diagnosis. Panel (B) Prevalence of ≥1 outpatient visit in the 7 days prior to hospitalization stratified by age and sex. Panel (C) Total number of outpatient encounters for study cohort within the 7 days leading up to CVT admission. Panel (D) Number of inpatient encounters for study cohort within the 7 days leading up to CVT admission. Note: Encounters within 1 or 2 days prior to CVT admission in Panel (D) are combined due to privacy requirements for small values.
Within the 7 days prior to hospital admission with incident CVT, 31/554 (5.6%) of this cohort had ≥1 inpatient health care encounter during a previous hospitalization without receiving a diagnosis of CVT. Within 3 days of admission with incident CVT, 11/554 (2.0%) had an inpatient health care encounter. The three most common categories for hospitalization were related to pregnancy, infectious causes (both systemic and intra-cranial infections) followed by non-infectious central nervous system causes.
Factors associated with presence of outpatient encounters
Frequency of outpatient healthcare encounters by age and sex is shown in Figure 1, Panel B. Compared to males, females more frequently had ≥1 outpatient encounter in both the 7 and 3 days leading up to their CVT diagnosis (7 days: 64.8% vs. 49.4%, p < 0.001; 3 days: 51.8% vs. 40.5%, p = 0.01). This difference remained significant after removal of peripartum and traumatic CVTs (n = 83). Mean age was lower in individuals with ≥1 prior outpatient encounter 7 days prior to CVT admission compared to those who did not. However, this difference was not statistically significant (Table 1). Patients with traumatic CVTs were less likely to have a pre-admission healthcare interaction either at 7 or 3 days (7 days: 37.3% vs 60.0 % p = 0.003; 3 days: 23.5% vs. 49.1%, p = 0.001). There were no significant differences in frequency of outpatient encounters by cancer or peripartum status (Table 1).
Multivariable logistic regression modelling showed that females had higher odds of having ≥1 outpatient health encounter within the 7 and 3 days prior to admission compared to men (OR = 1.79; 95% CI: 1.25, 2.57) after adjusting for age, time epoch of admission, peripartum status, traumatic mechanism for CVT and recent cancer diagnosis (Table 2). After controlling for sex, time epoch of admission, peripartum status and CVT risk factors, all older age groups had lower odds of having a previous outpatient encounter compared to patients under age 35 (Table 2). This was significant for all age groups for encounters within 3 days and significant for outpatient encounters within 7 days except patients aged≥65. There were no significant interactions between age and sex. After controlling for age, sex and study epoch within the multi-variable regression, patients with traumatic CVTs were much less likely to have a pre-admission healthcare interaction both at 7 or 3 days. The direction of effect seen for age and sex was not altered by removing encounters in the day prior to admission but some categories were no longer significant due to smaller numbers (Table 3).
Table 2. Multivariate logistic regression for associations with odds of having an outpatient healthcare encounter within the 7 and 3 days prior to cerebral venous thrombosis (CVT) diagnosis
Table 3. Sensitivity analysis with exclusion of encounters one day prior to cerebral venous thrombosis (CVT) diagnosis
Physician specialty for encounters in the 7 days prior to CVT admission
Most outpatient encounters in the 7 days prior to CVT diagnosis were with general practice (58.0% of encounters). The next most common specialties for outpatient encounters within 7 days prior to CVT admission were emergency medicine (8.3%), neurology (5.7%), internal medicine (4.6%), and ophthalmology (3.3%) (Table 4). There were no sex differences in presentation to emergency medicine vs. general practice. The most common provider specialties remained unchanged after encounters from the day prior to CVT admission were removed (numbers provided in brackets in Table 4).
Table 4. Physician specialties for healthcare encounters in the 7 days prior to cerebral venous thrombosis (CVT) diagnosis for the study cohort
Values in brackets represent results from the sensitivity analysis (i.e., following removal of services provided on the day prior to CVT admission).
For inpatient encounters in the 7 days prior to CVT diagnosis, the most common associated specialty was also general practice (34.9%), internal medicine (16.0%), and obstetrics and gynecology (11.3%) (Table 4).
Associations with CVT complications, mortality and length of stay
There were no differences in the deaths during index hospitalization for those with vs. without encounters within 7 days nor for those with vs. without encounters within 3 days. Inpatient length of stay also did not differ between the groups for encounters within 7 days and for 3 days), nor did the presence of intracranial hemorrhage or seizures (Table 1).
Discussion
In this population-based study of patients hospitalized with incident CVT in BC, Canada, there was a high proportion (57.9%) of individuals who had ≥1 outpatient physician encounter within the 7 days prior to their CVT hospitalization. We found females and younger patients had a higher odds of having an outpatient physician encounter in the 7 days prior to their CVT admission after adjustment for common CVT mechanisms (peripartum status, recent cancer diagnosis, head trauma), and time period of hospitalization. These findings suggest that a proportion of those admitted to hospital with CVT, particularly women and younger adults, likely did not receive a correct diagnosis for CVT-related symptoms that prompted prior presentations to medical attention in the days before. A comparatively small proportion (5.3% at 7 days, 2.0% at 3 days prior to CVT hospitalization) had an inpatient encounter with subsequent discharge prior to readmission with CVT. It is possible that some of these encounters were in part related to medical conditions that may have led to a subsequent CVT (i.e., head or neck infection, birth); others may potentially represent admissions for symptoms related to a CVT (e.g., seizure, headache, nausea or vomiting) that was not diagnosed until a subsequent admission.
Studying pre-hospitalization care patterns may allow for future opportunities for the earlier diagnosis of diseases that worsen in severity over days to weeks. We found that most individuals had at least one outpatient physician encounter within the 7 days prior to their CVT hospitalization. By comparison, a recent systematic review showed that 10.3% to 52.9% of patients admitted to hospital with sepsis had outpatient healthcare encounters prior to hospitalization. Reference Flannery, Venn and Gusovsky15 A previous study of emergency room visits using administrative data from New York, California, and Florida from 2005-2013 showed that 3.6% of patients with CVT had an emergency room visit in the 14 days before admission Reference Liberman, Gialdini, Bakradze, Chatterjee, Kamel and Merkler1 for headache or seizure but did not comment on the overall burden of emergency room visits. Within our study, 8.3% of prior outpatient encounters were seen by an emergency medicine physician within 7 days prior to CVT admission. In the emergency department setting, women with stroke are more often misdiagnosed compared to men. Reference Newman-Toker, Moy, Valente, Coffey and Hines6,Reference Tarnutzer, Lee, Robinson, Wang, Edlow and Newman-Toker17 Gender differences in health services utilization between men and women in Canada have previously been described and show that women have higher healthcare utilization but are only half as likely to use the emergency room compared to men. Reference Kazanjian, Morettin and Cho16 Although healthcare utilization prior to CVT was higher in women in our study, we did not find a difference between men and women with regards to visits with general practitioners vs. emergency physicians. However, it is possible that some general practitioner visits actually represent presentations to the emergency room with a GP-emergency provider, as most emergency rooms in BC employ a proportion of GPs with additional emergency medicine training. Although it is highly likely that a proportion of GP encounters (the most common specialty for pre-hospitalization encounters) in fact represent emergency room assessments, it is a limitation of our study that we are unable to confirm this by distinguishing which general practice encounters were in a community-based clinic as compared to the emergency room.
Our findings differ from a sub study of the International Study on Cerebral Vein and Dural Sinus Thrombosis (ISCVT), a large prospective cohort of patients with CVT, in which delays to diagnosis were more common in men. Reference Ferro, Canhão and Stam18 However, the nature of this study’s cohort is different, given that participants in ISCVT were consecutively recruited from academic hospitals with a primary diagnosis of symptomatic CVT and 74.5% of participants in that cohort were female whereas our study used an unselected population-based cohort where 55.4% are female. The female-to-male ratio among CVT patients in more recent population-based cohorts Reference Nasreen, Calzavara and Sundaram19–Reference Ruuskanen, Kytö, Posti, Rautava and Sipilä21 has been lower than what has been previously reported from prospective registries, suggesting differences in patient populations captured resulting from different approaches to ascertainment.
A limitation of the study is that our cohort did not include linkages to outpatient diagnostic codes nor diagnoses from emergency room visits through the National Ambulatory Care Reporting System which prevented us from being able to distinguish encounters that were most likely related to the CVT vs. those for clearly unrelated symptoms. Unlike a previous US study, Reference Liberman, Gialdini, Bakradze, Chatterjee, Kamel and Merkler1 we did not find a difference in length of stay (or any other outcome measure) in CVT patients with and without pre-hospitalization healthcare encounters. However, being limited to administrative data, detailed information regarding markers of disease severity, such as depressed level of consciousness on admission, radiological features or longer-term functional outcomes such as modified Rankin Scale scores and patient-centered outcomes common after CVT including such as headache or cognitive changes Reference Preter, Tzourio, Ameri and Bousser23,Reference Ji, Zhou and Wu24 are not available. Related to this, using administrative data, we are unable to comment on the exact delay from symptom onset to eventual CVT diagnosis.
Another limitation of our study is the lack of data on referrals and imaging generated from outpatient visits. Future work examining referrals for specialty assessment and imaging would allow for quantification for effects of interventions that reduce waits in emergent access to specialists and outpatient imaging. In conditions such as symptomatic carotid stenosis, previous Canadian research has demonstrated that the location of initial health care contact (ambulatory clinic vs. emergency room) was a significant factor in reducing time to imaging and subsequent treatment. Reference Charbonneau, Bonaventure, Drudi, Beaudoin, Blair and Elkouri22 Finally, CVT cases were identified using ICD-9 and 10 codes which although previously validated, still show lower sensitivity than other methods such as radiology reports or EMR free-text searches. Reference Zhou, Yu, Hall, Hill and Field10
Unlike other causes of stroke, symptoms in CVT can be non-focal and with an insidious symptom onset. Within our Canadian population-based CVT cohort across almost 2 decades, over half of patients had a healthcare encounter within 7 days of their hospital admission with incident CVT, mostly with general practitioners. Outpatient encounters were more common in women compared to men and in younger patients. From our study, it is uncertain if this is due to differences in health-seeking behavior, intrinsic differences in CVT symptoms or differences in how symptoms in younger people and women are interpreted by physicians. It is likely that a proportion of these encounters may represent a missed opportunity for a more timely diagnosis. Further education about CVT, its presenting symptoms, and at-risk populations should be provided to front-line practitioners assessing patients with undifferentiated symptoms to redirect them for urgent imaging. Public education campaigns bringing awareness to biases faced by women and young adults with stroke, which have been put forth by the Heart and Stroke Foundation of Canada and other organizations, should continue to include CVT survivor stories to encourage self-advocacy from patients.
Acknowledgements
This project is supported by a project grant from the Canadian Institutes of Health Research (RN 387091 - 420683).
Author contributions
ZW organized and analyzed the data and prepared the first draft of the manuscript. TSF was involved in the conception of the project and participated in editing and revising the manuscript. LWZ was also involved in the conception of the project, provided oversight of the data analysis, participated in the preparation of the first draft of the manuscript, and provided edits and revisions to the manuscript.
Funding statement
ZW: Financial support from CIHR Project Grant (RN 387091 - 420683). TSF: Support from CIHR Project Grant (RN 387091 - 420683) and Sauder Family/Heart & Stroke Professorship of Stroke Research; Consulting fees from Bayer Canada, AstraZeneca, Roche Canada; Honoraria from HLS Therapeutics; Payment for expert witness testimony; Board Member of DESTINE Canada. LWZ: Fellowship salary support from CIHR Project Grant (RN 387091 - 420683) and salary support from Vancouver Coastal Health Research Institute.
Competing interests
The authors report no disclosures relevant to the manuscript.
