Estimation of the prevalence of diagnosed diabetes from primary care and secondary care source data:

1
Estimation of the prevalence of diagnosed
diabetes from primary care and secondary care
source data comparison of record linkage with
capture-recapture analysis

• J. N. Harvey, L. Craney, D. Kelly
• Group members
• Suzette Poliquin, Eduardo da Silveira, Eric Lam,
  Susie Lau

2
Background Information

• Objectives
• 1. Compare GP records and hospital records with
  capture-recapture analysis in determining the
  prevalence of type 1 and type 2 diabetes in a UK
  white population (population based surveillance)
• 2. Assess whole population trends in diabetes
  prevalence and treatment by comparison with
  previous studies

3
Target Population

• Total population from county of Clwyd, North
  Wales, defined by postal code and resident as of
  March 1998 ? total population 418,200 people
• gt 99 white

4
Method of Sampling

• Data obtained from
• A) 3 District General Hospitals
• Patient Administration Computer Systems (PAS)
  patients coded as having diabetes (ICD 9 code
  250)
• Hospital diabetes clinics -type of diabetes coded
  at this stage
• Diabetes nurses records of patients having
  attended diabetic clinic in previous 15 years

5
Method of Sampling Contd

• B) 74 Practices (Primary Care)
• list of diabetic patients was supplied
• reference date - March 1998 must be resident in
  catchment area
• Case Definition
• Type 1 diabetes diagnosis before age 40, on
  insulin treatment within 1 year from diagnosis
  (WHO criteria)
• Exclusions
• Intermediate glucose intolerance and gestational
  diabetes

6
Data Base Validation

• One year after starting the study, further
  inquiries made of all patients whose name did not
  appear in more than one source or type of
  diabetes and treatment not listed
• Used hospital and primary care records and the
  NHS Wales administrative Register to eliminate
  those who did not have diabetes, had died or were
  not resident in the catchment area on the
  reference date of March 1998

7
Capture - recapture Analysis

• Two source model used
• Hospital source combination of PAS system,
  hospital clinics, diabetes nurses records
• Primary Care Source combined lists from the
  general practitioners

8
Capture-recapture Contd

• Predicted number of cases (N) calculated from a
  formula (LaPorte) to give the overall unadjusted
  prevalence for all patients.
• Age adjusted prevalence rates calculated for each
  five year age by gender and diabetes type

9
Critique

• 1. Study design and sampling method appropriate
  for the research question?
• -Yes, however missing patients referred to
  tertiary care, homeless people
• 2. Sampling frame appropriate?
• -Yes for comparing source linkage and
  capture-recapture analysis
• - No for establishing trends

10
Critique

• 3. Sample size adequate?
• -Yes for county Clwyd
• -Representativeness of North Wales?
• 4. Objectives, suitable and standard criteria
  used for measurement of outcome?
• - WHO criteria Type 1 diabetes diagnosis
  before age 40, on insulin treatment within 1 year
  from diagnosis
• -Over estimation of type 1 diabetics

11
Critique

• 5. Outcome measured in an unbiased fashion?
• -Information bias, missing type 2 diabetics not
  having consulted a physician or not treated in
  hospital
• 6. Is the response rate adequate? Refusers
  described?
• -Response rate adequate (8,877)
• -Refusers were considered to be the
  unclassified(301), no further description given

12
Critique

• 7. Estimates of prevalence or incidence given
  with CI in detail by subgroups?
• Using the capture-recapture methodology, age
  specific prevalence for type 1 and type 2
  diabetes was given by gender, with the 95
  confidence intervals .

13
Figure 1 Age and gender specific prevalence of
type 1 diabetes calculated by capture-recapture
analysis. Error bars represent 95 confidence
intervals. Harvey et al, Journal of Epidemiology
and Community Health 2002 5618-23
14
Figure 2 Age and gender specific prevalence of
type 2 diabetes calculated by capture-recapture
analysis. Error bars represent 95 confidence
intervals Harvey et al, Journal of Epidemiology
and Community Health 20025618-23
15
Critique

• 8. Implications of results for action or
  intervention
• Effective monitoring is important for
  development of disease prevention programs
  allocation of health-care resources
• i.e. Type 1 diabetes in men between age 35 and
  45
• Type 2 diabetes in men after age 45