Patterns of health seeking behaviour and illness in families where children go into public care

1
Patterns of health seeking behaviour and illness
in families where children go into public care

• Doug Simkiss
• Margaret Thorogood, Nigel Stallard,
• Nick Spencer

2
Introduction

• Study hypothesis
• Methodology
• Results to date
• Discussion

3
Study hypothesis

• Families (specifically mother - child pairs) with
  children taken into public care have different
  patterns of illness and health service use before
  the children are removed into care compared to
  other families.

4
Methodology - GPRD

• Largest database of anonymous primary care
  information in the world.
• 400 GP surgeries in UK
• 3 million patients (39 million patient years)
• 250,000 family groupings
• Can identify when children leave and parent
  stays, key word search for foster, in care.

5
Cases

• Mother child pairs where the child has a Read /
  OXMIS code for care. The oldest child in the
  family is the index child and must be 1-15 years
  of age at the time of the index event. Only pairs
  with 12 months up to standard data prior to
  the event will be included.

6
Controls

• Mother-child pairs (4 control pairs per case
  pair) on the same practice register who are not
  cases. Control children are matched on year of
  birth and gender and only controls with 12
  months of up to standard data prior to the index
  event of the case child will be included.

7
GPRD data

• 2954 cases identified on database with 11,816
  controls
• Caveats ? 351 cases and 1363 controls
• GPRD have also given data on
• Entire sib ship with a code for care and birth
  order.
• Children with a code for care who are lt 1 year
  old at index event

8
Data validation questionnaire
9
Data analysis

• Concentrated on clinical events, therapy and
  referral files
• All clinical events entries coded by two
  independent researchers and then discussed.
• Therapy files coded for named drugs
• Referral files coded for nature of referral

10
Data analysis mothers clinical events
11
Data analysis childs clinical events
12
Results

• 74 variables created
• 42 significantly different between cases and
  controls
• 2 conditional logistic regression models
• Sensitivity the percentage of cases who have a
  particular variable
• Specificity the percentage of controls who do
  not have a particular variable

13
Results

• Positive predictive value the percentage of
  those with a positive result for a variable who
  are truly cases
• Negative predictive value the percentage of
  those with a negative result for a variable who
  are truly controls

14
Child drug misuse

• Kid_drug_misuse p 0.0000005
• Sensitivity 9 / (9342) 3
• Specificity 1363 / (1363 0) 100
• Positive predictive value 9 / (90) 100
• Negative predictive value 1363 / (1363342) 80

15
(No Transcript)
16
Variable 2 by 2 tables

• Sensitivity 23
• Specificity 95
• PPV 54
• NPV 83

• Sensitivity 14
• Specificity 99
• PPV 96
• NPV 82

17
(No Transcript)
18
Variable 2 by 2 tables

• Sensitivity 27
• Specificity 97
• PPV 69
• NPV 84

• Sensitivity 51
• Specificity 80
• PPV 40
• NPV 86

19
Variable 2 by 2 tables

• Sensitivity 21
• Specificity 95
• PPV 50
• NPV 82

• Sensitivity 6
• Specificity 99
• PPV 57
• NPV 80

20
Variable 2 by 2 tables

• Sensitivity 27
• Specificity 64
• PPV 16
• NPV 77

• Sensitivity 4
• Specificity 99
• PPV 81
• NPV 80

21
Variable 2 by 2 tables

• Sensitivity 38
• Specificity 81
• PPV 34
• NPV 84

• Sensitivity 1
• Specificity 99
• PPV 80
• NPV 80

22
Variable 2 by 2 tables

• Sensitivity 21
• Specificity 95
• PPV 50
• NPV 82

• Sensitivity 19
• Specificity 87
• PPV 38
• NPV 81

23
Discussion

• Mobility
• Socioeconomic status

24
Mobility

• Friedlaender et al. Patterns of health care use
  that may identify young children who are at risk
  for maltreatment Pediatrics. 2005 116 (6) 1303
  8.
• In essence the only identified difference was
  increased mobility in primary care provider in
  the maltreated group.

25
Socio - economic status

• SES scores for practices
• Awaiting SES scores based on enumerator district