Concept: Stroke - Measuring Prevalence

Definition of Stroke

A stroke occurs when there is a sudden death of brain cells due to a lack of oxygen when the blood flow to the brain is impaired by blockage or rupture of an artery to the brain. Symptoms of a stroke depend on the area of the brain affected. The most common symptom is weakness or paralysis of one side of the body with partial or complete loss of voluntary movement or sensation in a leg or arm. Other common symptoms include speech problems, weak facial muscles, numbness and tingling. A stroke involving the base of the brain can affect balance, vision, swallowing, breathing and consciousness.

A stroke is also known as Cerebral Vascular Accident (CVA). See MedlinePlus® - Health Topics - Stroke for more information.

Literature Review

Table 1 summarizes seven studies, published prior to 2006, that use administrative data to ascertain cases of stroke. A significant issue in the research has been the choice of diagnostic codes to identify disease cases. Some studies adopted the broadest possible set of codes, which included ICD-9-CM 430 to 438 (cerebrovascular disease). However, others excluded specific codes, like 437, which represent stroke of undetermined causes. The limitation of all of the studies identified in Table 1 is that they only relied on hospital separations to identify stroke cases.

Manitoba Stroke Algorithms

The following sections describe stroke algorithms defined in research at the Manitoba Centre for Health Policy (MCHP).

DATA SOURCES:

In the following algorithms, two common databases are used in identifying stroke:


• hospitalization / hospital separation information is found in the Hospital Abstracts data; and
  
• cause of death information is found in the Vital Statistics Mortality Registry data.

Some MCHP research has involved other databases and these are described in each specific research project below.

1. Lix et al. (2006)

In Lix et al.(2006) , both sensitive and specific sets of algorithms were investigated using broad (ICD-9-CM 430-438) and narrow (ICD-9-CM 430, 431, 434, 435, 436) sets of diagnostic codes, respectively. These diagnostic codes were identified in hospital discharge abstracts and medical services/physician claims data.

The following drug categories were used to identify prescription drug records for inclusion in the research: anti-platelet agents such as aspirin (ASA) at 81 or 325 mg once a day, clopidogrel, ticlopidine, dipyridamole, and combination agents such as Aggrenox (ASA 25mg dipyridamole 200mg slow release) and oral anti-coagulants such as warfarin, pheninidione, and nicoumalone. The fifth level Anatomical Therapeutic Chemical (ATC) codes selected for the research were B01AA02, B01AA03, B01AA07, B01AC07, B01AB01, B01AC30, B01AC05, B01AC06, B01AC04, B01AB09, B01AB04, B01AB10. Thrombolytic agents such as rt-PA (recombinant tissue plasminogen activator) and intravenous anti-platelet agents (anti GP 2b/3a) such as abciximab, tirofiban and eptifibatide are markers for stroke therapy when administered on an inpatient basis. However, they can not be used as markers of stroke therapy when used on an outpatient basis. Therefore prescription drug records with Drug Identification Numbers (DINs) for these drugs were not included in the algorithms.

Table 2 reports sensitivity, specificity, kappa, Youden's index, PPV (positive predictive value) and NPV (negative predictive value) for each of the investigated algorithms. The validation was conducted using self-report of stroke (i.e., Do you suffer from the effects of a stroke) in the Canadian Community Health Survey cycle 1.1. The validation was based on 108 cases. It is important to note that none of the algorithms relied solely on prescription drug records to identify non-fatal stroke cases. This is because the drugs selected are not used exclusively as markers for stroke.

Calculating Population-based Prevalence Rates

The population registry was used by Lix et al. (2006) to define population cohorts to derive numerator and denominator data for calculating crude provincial prevalence estimates for each algorithm. Note: The Lix et al. (2006) estimates are based on the population 19 years of age and older.

2. Lix et al. (2008)

Lix et al. (2008) provided an update to the 2006 study with a report titled Defining and Validating Chronic Disease: An Administrative Data Approach. An Update with ICD-10-CA. The purpose of the 2008 report is to examine the validity of administrative data for monitoring the prevalence of chronic disease in Manitoba. Specific objectives are:


• Report relevant ICD-10-CA codes for ascertaining cases of chronic disease in administrative health data;
  
• Evaluate the validity of multiple algorithms for identifying disease cases from Manitoba administrative data.

The 2008 report uses the same methods and some of the same algorithms as described in the 2006 report, with the following modifications:


• ICD-10-CA codes were used to define specific chronic diseases from hospital separation data, beginning April 1, 2004. This is due to a change in coding systems used in Manitoba hospitals. Only the broad (full) set of ICD-9-CM codes (430-438) identified in the 2006 report were used to identify stroke in hospital cases prior to April 1, 2004, and in the physician claims data.
  
• data from the Canadian Community Health Survey (CCHS), cycle 3.1, collected from January 2005 to January 2006 were used to evaluate the validity of the administrative data. The cohort consisted of 5,099 adults 19+ years of age.
  
• additional algorithms were developed for the 2008 report.

The following ICD-10-CA codes were used to define stroke in administrative hospital separation data from April 1, 2004 to March 31, 2006:


• I60 - I69: cerebrovascular diseases

Table 3 contains the estimates of agreement kappa(κ), sensitivity, specificity, Youden's Index, PPV (positive predictive value) and NPV (negative predictive value) for each of the 24 algorithms that were investigated for stroke.

Discussion of the validation results for stroke can be found in the full report: Chapter 8: Stroke

Calculating Population-based Prevalence Rates

Crude provincial prevalence estimates for the 24 stroke algorithms are reported in Table 4. These estimates are based on the population 19 years of age and older.

Discussion of the prevalence rates for stroke can be found in the full report: Chapter 8: Stroke

3. Fransoo et al. (2009, 2011, 2013, 2019), Martens et al. (2010) and Chartier et al. (2012, 2015)

In the following deliverables:


• Manitoba RHA Indicators Atlas 2009 by Fransoo et al. (2009),
  
• Profile of Metis Health Status and Healthcare Utilization in Manitoba: A Population-Based Study by Martens et al. (2010),
  
• Adult Obesity in Manitoba: Prevalence, Associations, and Outcomes by Fransoo et al. (2011),
  
• Health and Healthcare Utilization of Francophones in Manitoba by Chartier et al. (2012), and
  
• The 2013 RHA Indicators Atlas by Fransoo et al. (2013);
  
• Care of Manitobans Living with Chronic Kidney Disease by Chartier et al. (2015); and
  
• The 2019 RHA Indicators Atlas by Fransoo et al. (2019);

... stroke was defined as:


• an inpatient hospitalization with the most responsible diagnosis of stroke: ICD-9-CM codes 431, 434, or 436 or ICD-10-CA codes I61, I63, I64; and a length of stay of one or more days (unless the patient died in hospital); OR
  
• a death with stroke listed as the cause of death on the Vital Statistics death record.

This definition is usually restricted to Manitoba residents age 40 or older.

Note: This definition will not capture minor strokes, which did not result in hospitalization or death.

Further Information

For more information on how stroke / cerebral vascular accidents (CVA) was investigated in each of these research projects, please read the related sections in the following reports:


• Manitoba RHA Indicators Atlas 2009 (2009) deliverable - 4.8 Stroke Rates
  
• Profile of Metis Health Status and Healthcare Utilization in Manitoba: A Population-Based Study (2010) deliverable - 5.10 Stroke Incidence Rates
  
• Adult Obesity in Manitoba: Prevalence, Associations, and Outcomes (2011) deliverable - Stroke
  
• Health and Healthcare Utilization of Francophones in Manitoba (2012) deliverable - 5.4 Stroke
  
• The 2013 RHA Indicators Atlas (2013) deliverable - 4.12 Stroke Rates
  
• in the Care of Manitobans Living with Chronic Kidney Disease by Chartier et al. (2015), they investigated the prevalence and relative risk of stroke as a comorbidity to end stage kidney disease (ESKD) and chronic kidney disease (CKD). For more information on prevalence and relative risk, see:
  
  • Table 4.1 Physical Health of Adults in End Stage Kidney Disease; and
    
  • Table 4.5 Physical Health of Adults in Chronic Kidney Disease.
• The 2019 RHA Indicators Atlas (2019) deliverable - 4.14 Stroke Rates

4. Finlayson et al. (2010)

Finlayson et al. (2010) define stroke as one or more hospitalizations OR one or more physician visits over a five-year time period for those aged 19+ where the events are coded with an ICD code representing stroke (ICD-9-CM codes 430-438).

5. Martens et al. (2015)

In The Cost of Smoking: A Manitoba Study deliverable by Martens et al. (2015) they calculated stroke prevalence rates two ways; one using administrative data and the other using self-reported survey data. Using the administrative data, the weighted crude prevalence of stroke was calculated for survey respondents aged 12 and older in the five years prior to their survey date. Stroke history was defined by one of the following conditions:


• one or more hospitalizations with a diagnosis of stroke, ICD-9-CM codes 430-438; ICD-10-CA codes I60-I69; or
  
• two or more physician visits with a diagnosis of stroke (ICD-9-CM codes as above).

Using the survey data, a question about stroke was asked in all the survey waves with the exception of the Canadian Community Health Survey (CCHS) 2.2. In the Manitoba Heart Health Survey respondents were asked, "Have you ever had a stroke?" Possible responses include "yes", "no" or "not sure". In the National Population Health Survey (NPHS) and CCHS respondents were asked, "Do you suffer from the effects of a stroke?" Possible responses include "yes", "no" or "don't know." The weighted crude self-reported prevalence of stroke was calculated for survey respondents aged 12 and older as the percentage of respondents who answered "yes" out of all respondents who gave a valid answer. Respondents who answered "don't know" or "not sure", and those with missing or invalid data were excluded from the prevalence calculation.

For more information, Table 4.8 Chronic Diseases of Estimated-Population-Based Sample* at Time of Survey by Smoking Status Categories lists the prevalence rates of stroke from survey (self-reported) and administrative data found in this report.