1. The 2002 Healthcare Conference
abcd
The 2002 Healthcare Conference
29 September-1 October 2002
Scarman House, The University of Warwick, Coventry

2. Session B1 : Critical Illness
Trends in Critical Illness
Heart Attack & Stroke
Working Party / Research Sub-group Report
Scott Reid & Joanne Wells

3. Critical Illness Trends Working Party
Our Aims :
To examine underlying trends in the factors influencing UK Insured Critical Illness claim rates, and from these, to assess :
The historic trend in incidence and death rates for the major CI’s
Any pointers for future trends in Standalone CI, Mortality and hence Accelerated CI.
Formed in March 2001

4. Actuaries Medical Expert Scott Reid Joanne Wells
Sub-Group Members
Actuaries
Scott Reid
Joanne Wells
Medical Expert
Dr Richard Croxson - Consultant Cardiologist

5. Contents Slide
Background and set the scene
Variations by deprivation category
Variations by ICD code
Impact of smoking
Future influences on rates
Project population incidence rates

6. Contents Slide
Background and set the scene
Variations by deprivation category
Variations by ICD code
Impact of smoking
Future influences on rates
Project population incidence rates

7. Background and set the scene
Re-cap of previous work
Data sources used
Scottish
English
Next step forward

8. Background and set the scene
Re-cap of previous work
Data sources used
Scottish
English
Next step forward

9. Re-cap of previous work
Population trends – Scotland and England
Heart attack
Stroke
CABG
Angioplasty
Broad brush analysis of smoker prevalence

10. Population trends – Scotland and England
Heart attack
Significant mortality and incidence improvements
Scottish rates at a significantly higher level
Stroke
English data unclear
Scottish data
Flat trend during 1980’s
Deterioration during early 1990’s

11. Broad brush analysis of smoker prevalence
Smoking is a key risk factor
Reduction in smoking prevalence
Scottish and English smoker prevalence patterns
Scottish trends

12. Background and set the scene
Re-cap of previous work
Data sources used
Scottish
English
Next step forward

13. Scottish population – ISD data
Data sources used
Scottish population – ISD data
Good quality
Patient based
English population - HES data
Data quality is questionable
Episode based

14. Background and set the scene
Next step
Insured trends
Understanding the main drivers to cause trends
Smoker differentiated rates
Future influences
Overall trend pattern

15. Contents Slide
Background and set the scene
Variations by deprivation category
Variations by ICD code
Impact of smoking
Future influences on rates
Project population incidence rates

16. Variations by deprivation class
Why split by deprivation class?
Data sources used
Explanation of deprivation scores and categories
Overall trends by gender
Heart attack
Stroke
Conclusion

17. Variations by deprivation class
Why split by deprivation class?
Data sources used
Explanation of deprivation scores and categories
Overall trends by gender
Heart attack
Stroke
Conclusion

18. Why split by deprivation class?
To understand trend at insured level
Regional variations and Target market variations
Understand the main drivers of health inequalities
Black report 1980: “..the main influence on the inequalities in health which were observed lay in the material circumstances in which people live”
Deprivation and Health in Scotland, 1991 (Carstairs & Morris)
Classification by postcode; overcomes the weakness of Occupational classification.

19. Variations by deprivation class
Why split by deprivation class?
Data sources used
Explanation of deprivation scores and categories
Overall trends by gender
Heart attack
Stroke
Conclusion

20. Data sources used Incidence data Mortality data Population data
SMR1/01, Information Statistics Division NHS Scotland
General Registers Office for Scotland
Mortality data
Population data
1981 Population Census
1991 Population Census
Split by
CI condition
ICD code
Gender
5 year age bands
deprivation category

21. Variations by deprivation class
Why split by deprivation class?
Data sources used
Explanation of deprivation scores and categories
Overall trends by gender
Heart attack
Stroke
Conclusion

22. Explanation of deprivation scores and categories
Carstairs & Morris 1991 deprivation categories
Four indicators – to derive a composite score
Overcrowding
Male unemployment
Low social class
No car
Deprivation score divided into 7 separate categories
1 – the most affluent group ……
7 – the most deprived group

23. Explanation of deprivation scores and categories

24. Explanation of deprivation scores and categories

25. Variations by deprivation class
Why split by deprivation class?
Data sources used
Explanation of deprivation scores and categories
Overall trends by gender
Heart attack
Stroke
Conclusion

26. Trends in incidence of first heart attack for males in Scotland, as a % of 1981 Value, 1981 to 2000

27. Trends in incidence of first heart attack for males in Scotland, as a % of 1981 Value, 1981 to 2000

28. Trends in incidence of first heart attack for males in Scotland, as a % of 1981 Value, 1981 to 2000

29. Trends in incidence of first heart attack for males in Scotland, per of population, 1981 to 2000

30. Trends in incidence rate of first heart attack for males aged 40 to 64 in Scotland, per of Population, 1981 to 2000

31. Trends in incidence rate of first heart attack for males aged 40 to 64 in Scotland, as a % of 1986 Value, 1981 to 2000

32. Trends in incidence rate of first heart attack for females aged 40 to 64 in Scotland, per of Population, 1981 to 2000

33. Trends in incidence rate of first heart attack for females aged 40 to 64 in Scotland, as a % of 1986 Value, 1981 to 2000

34. Summary of heart attack trends by deprivation class

35. Summary of heart attack trends by deprivation class, Males aged 40 to 64, 1986 to 2000

36. Summary of heart attack trends by deprivation class, Males aged 40 to 64, 1986 to 2000

37. Brief interpretation of heart attack trends
Male
Female
Mortality
Positive correlation between affluent and deprived groups
Incidence
Less clear. Weak negative correlation where deprived group has higher improvement
Postive correlation between affluent and deprived groups except for categories 6 and 7

38. Variations by deprivation class
Why split by deprivation class?
Data sources used
Explanation of deprivation scores and categories
Overall trends by gender
Heart attack
Stroke
Conclusion

39. Trends in incidence rate of first stroke for males aged 40 to 64 in Scotland, as a % of 1981 value, 1981 to 2000

40. Trends in incidence rate of first stroke for males aged 40 to 64 in Scotland, as a % of 1981 value, 1981 to 2000

41. Trends in incidence rate of first stroke for males aged 40 to 64 in Scotland, as a % of 1981 value, 1981 to 2000

42. Trends in incidence rate of first stroke for males aged 40 to 64 in Scotland, per of Population, 1981 to 2000

43. Trends in incidence rate of first stroke for males aged 40 to 64 in Scotland, per of Population, 1986 to 2000

44. Trends in incidence rate of first stroke for females aged 40 to 64 in Scotland, per of Population, 1986 to 2000

45. Trends in mortality rate by stroke for males aged 40 to 64 in Scotland, per of Population, 1986 to 2000

46. Trends in mortality rate by stroke for females aged 40 to 64 in Scotland, per of Population, 1986 to 2000

47. Summary of stroke trends by deprivation class

48. Summary of stroke trends by deprivation class, Males aged 40 to 64, 1986 to 2000

49. Summary of stroke trends by deprivation class, Females aged 40 to 64, 1986 to 2000

50. Brief interpretation of stroke trends
Male
Female
Mortality
Positive correlation between affluent and deprived groups
Weak negative correlation between affluent and deprived groups
Incidence
Weak positive correlation between affluent and deprived groups
Weak postive correlation between affluent and deprived groups

51. Contents Slide
Background and set the scene
Variations by deprivation category
Variations by ICD code
Impact of smoking
Future influences on rates
Project population incidence rates

52. Variation by ICD code Why? Heart attack Stroke
Understanding of overall rate
Explain which components have influenced overall trend
Heart attack
ICD9 code 410
Unstable angina ICD9 code 413
Stroke
ICD9 codes 430 to 437 excluding 435

53. Variation by ICD code Why? Heart attack Stroke
Understanding of overall rate
Explain which components have influenced overall trend
Heart attack
ICD9 code 410
Unstable angina ICD9 code 413
Stroke
ICD9 codes 430 to 437 excluding 435

54. Trend in first incidence rate for males aged 40 to 64 by ICD code 410 and 413, per of Population, 1981 to 2000

55. Trend in first incidence rate for females aged 40 to 64 by ICD code 410 and 413, per of Population, 1981 to 2000

56. Variation by ICD code Why? Heart attack Stroke
Understanding of overall rate
Explain which components have influenced overall trend
Heart attack
ICD9 code 410
Unstable angina ICD9 code 413
Stroke
ICD9 codes 430 to 437 excluding 435

57. Trend in first incidence rate for males aged 40 to 64 by ICD codes 430 to 437 excluding 435, per of Population, 1981 to 2000

58. Trend in first incidence rate for males aged 40 to 64 by ICD codes 430 to 437 excluding 435, 1981 to 2000

59. Trend in first incidence rate for females aged 40 to 64 by ICD code 430 to 437 excluding 435, per of Population, to 2000

60. Trend in first incidence rate for females aged 40 to 64 by ICD code 430 to 437 excluding 435, per of Population, to 2000

61. Summary of variation by ICD code
Heart attack:
ICD 410 is improving over time (Male 3.6% p.a., Female 3.3%)
ICD 413 is deteriorating (Male 5.2% p.a., Female 6.2% p.a.)
Troponins?
Stroke
Large component ICD 436 is improving over time (Male 5.2% p.a., Female 6.2%)
Remaining components ICD 430, 431,432,433, 434, 437 overall are deteriorating over time (Male 5.7% p.a., Female 4.3% p.a.)
Overall deterioration (Male 1.9%, Female 1.5%)
Overall flat trend over 1980’s, deterioration over 1990’s
Future?

62. Contents Slide
Background and set the scene
Variations by deprivation category
Variations by ICD code
Impact of smoking
Future influences on rates
Project population incidence rates

63. Contents Slide
Introduction
Variations by deprivation category
Variations by ICD code
Impact of smoking
Future influences on rates
Projecting population incidence rates

64. Impact of Smoking - “Ideal” Smoking Model
ix (pop) = ix(ns)* p(ns) + ix (s) * p(s) +ix(ex) * p(ex)
ix (pop) = ix(ns)* p(ns) + ix (s) * p(s) + t ix(ex t) * p(ex t)
…..

65. Impact of Smoking - Data Available
Epidemiological evidence on smoking
Case control studies, prospective cohort studies etc
Wide range of results
results become even more volatile if looking for age specific or duration smoked/quit specific results
often look at impact on mortality not incidence
cause investigated is often not an exact match e.g. coronary heart disease and not acute myocardial infarction
usually smoking status is only investigated at the start of the study period
Simplify and/or use proxies

66. Impact of Smoking - Data Available
Epidemiological evidence on smoking
Case control studies, prospective cohort studies etc
Wide range of results
results become even more volatile if looking for age specific or duration smoked/quit specific results
often look at impact on mortality not incidence
cause investigated is often not an exact match e.g. coronary heart disease and not acute myocardial infarction
usually smoking status is only investigated at the start of the study period
Simplify and/or use proxies

67. Impact of Smoking -Smoking Model
Age specific model
ix (pop) = ix(ns)* propx (ns) + ix (ns) * RRx *propx (s)
assumes that people move immediately from being a smoker to a never smoker
Ex-smoker model
ix(pop) = ix(ns) *propx(ns) + ix(ns) * RR1 *propx (ex1) +…
…….+ ix(ns) *RR(sm) propx(s)
assumes that the relative risks for smokers and ex-smokers are independent of age

68. Impact of Smoking - Males

70. Impact of Smoking - Females

73. Contents Slide
Introduction
Variations by deprivation category
Variations by ICD code
Impact of smoking
Future influences on rates
Projecting population incidence rates

74. Future Influences on Rates
Impact of changing risk factors on incidence
Focus on changes in diagnostic techniques and potential shocks in incidence
Discussion of impact of troponin on heart attack incidence
Interpreting trends in stroke incidence and the potential impact of brain imaging techniques

75. Troponin and Acute Coronary Syndrome
Number of medical papers available
Generally look at the change in the number of diagnoses in the spectrum of ACS

76. Spectrum of Acute Coronary Syndromes
K Fox., Heart; 2000;84;93-100

77. Troponin and Acute Coronary Syndrome
Number of medical papers available
Generally look at the change in the number of diagnoses in the spectrum of ACS
As might be expected sample sizes quite small
Definitions do not exactly match those used by the insurance industry
Average age is considerably higher than that of people who claim for critical illness

78. Troponin and Acute Coronary Syndrome

79. Troponin and Critical Illness
Medical studies give a range of results
Results need careful interpretation before trying to apply them to critical illness
More AMI being diagnosed but some an acceleration e.g.
subsequent heart attack
coronary artery bypass surgery

80. Troponin and Critical Illness
Percentage of hospitals where troponin is available
Scotland - 70% (Pell BMJ Vol 324)
England - 60% last year thought to be 70% to 80% this year
No clear consensus amongst cardiologists in the UK on the definition
New definition not disseminated until September 2000 so no effect on the data currently published

82. Stroke - Identifying Trends
“Stroke mortality is falling in many countries, but it is unclear whether this is due to a fall in stroke incidence, lower case fatality, or some artifact of the collection and analysis of routine mortality data.”
Stroke, A Practical Guide and Management. Warlow et al

83. Stroke - Identifying Trends
“In the few places where it has been measured reasonably reliably, stroke incidence seems to have declined, stayed the same, or increased. However it has been very difficult to use consistent methods and obtain large enough data sizes for precise estimates. In truth it is not very clear what incidence rates are doing.”
Stroke, A Practical Guide and Management. Warlow et al

84. Stroke - HES Data
Patients admitted to hospital generally have more severe strokes
The proportion of all strokes admitted is unknown and can change over time
Cannot identify first ever strokes
Double counting as patients move from one hospital service to another
Question mark over change in incidence coinciding with coding change

86. Will policyholders understand the differences?
Stroke - Diagnosis
ABI definition of stroke is different to that used by clinicians and to that used by the WHO
Consequences of, for example, increased MRI scanning could have a different impact under different definitions
Will policyholders understand the differences?
Milder strokes more often identified?
Patient expectations are rising
Diagnostically more competent
Helped by more sensitive brain imaging

87. Contents Slide
Introduction
Variations by deprivation category
Variations by ICD code
Impact of smoking
Future influences on rates
Projecting population incidence rates

88. Population - Projection of Incidence Rates