Sampling for Detailed NDIP Inspections Emma OKeefe StR in Dental - - PDF document

Slide 1

Sampling for Detailed NDIP Inspections

Emma O’Keefe StR in Dental Public Health Presented at NDIP Coordinators’ Study Day 26th July, 2010 (Updated 2014 in relation to sampling at CPP (Community Planning Partnership Level))

Please feel free to comment- I am not an expert, you have all been doing it for years- would welcome your comments suggestions but here is a start for sampling for detailed NDIP programme. Sampling advice- needs updating and in 2014 there was a further requirement to update this presentation as many areas planning at Community Planning Partnership level rather than Community Health Partnership level. This version is based on a presentation Gail Topping did a number of years ago and BASCD’s guidance document available on BASCD website. There is some conflicting advice, some out of date information so welcome to suggestions and comments as I am sure there are several ways of interpreting the info. BASCD Sampling Guidance paper- 2004 but before changes made to sample whole classes not individual children.

Slide 2

Aims & Objectives

• To give a working example of how to

carry out sampling at Board Level and CHP or CPP level Discuss Board Level, CHP or CPP level. Background information on Community Planning in Scotland WHAT IS COMMUNITY PLANNING? Community Planning is a process which helps public agencies to work together with the community to plan and deliver better services which make a real difference to people's lives. The aims of Community Planning in Scotland are:-  making sure people and communities are genuinely engaged in the decisions made

• n public services which affect them; allied to

 a commitment from organisations to work together, not apart, in providing better public services. There are two further key principles in addition to the two main aims outlined above: Community Planning as the key over-arching partnership framework helping to co-ordinate

• ther initiatives and partnerships and where necessary acting to rationalise and simplify a

cluttered landscape; the ability of Community Planning to improve the connection between national priorities and those at regional, local and neighbourhood levels. Board level- NDIP reports based on Now weighted results by deprivation in each NHS Board as measured by Scottish Index of Multiple Deprivation (SIMD). Some NHS Boards are interested in CHP or CPP levels to assist with local planning needs (see explanation above).

Slide 3

Why?

• Consistent
• Comparisons
• Opportunity to share experiences

Consistent approach to sampling- Scotland has both largest and smallest health board (or equivalent) in UK. Sampling is therefore difficult to standardise. Guidance out there but needs updating and clarification around a few points. Adequate numbers to allow meaningful comparisons between NHS Boards Sampling procedure differs from previous Scottish Health Boards Dental Epidemiological Programme in that whole classes are now selected to simplify the process for schools whilst ensuring that the results reflect the year population in Scotland.

Slide 4

Why sample?

• Resources
• Necessary
• Random and representative
• Sufficient size

Alternative is to inspect all children- not possible. Resources-  internal - Public Dental Service costs- staff time, costs, travel, get on with dental care  externa l- school curriculum, impact on staff and pupils Aim to examine sub set of children in year group who are representative of the whole year

• group. Sample must be random and representative and of sufficient size to give a

representative picture of dental health of the school year population.

Slide 5

Sampling protocol

• State maintained schools
• Minimum sample of 250 per area
• Select additional 15-20%
• All children in class
• Refresh sample each year

State, private, special needs Minimum 250 children or all children for smaller areas- sample size depends on level of precision required and what additional comparisons within the district are to be made. Minimum should be 250 children per area sampling at. There is mention in previous paperwork that a minimum of 20 schools or 8% of the population should be sampled. I sought clarification on this (in 2010) from Zoann Nugent who clarified that a minimum of 20 schools was suggested for urban areas in order to get a decent spread of schools geographically as in some urban areas you can pick up 250 kids in 3-4 schools. You need a sample that reflects urban/rural and large, medium and small schools. The 20 schools/8% are MINIMUM for large boards (e.g. NHSGGC and Lothian). 15-20% for absenteeism or declining, don’t substitute so looking at approximately 300 children in each area. All children in one class-basic and detailed at same time Composite-only relevant year children inspected Refresh randomisation each year Independent of any other work in schools Reserve schools Consent Communication- smooth running of programme- letter to head teachers, information consent letter to parents, letter to parents of child’s dental health and report to school

Slide 6

Sampling- Step 1 Board Area Worked example

• Lists from Local Education Dept.
• Exclusions
• 141 schools
• 3732 pupils
• 216 classes

Lists from Local Education Dept- most recent e.g. Current P6, use as proxy, check at start of new year number P7 (variation in/ex migration) or you can use present P7 year. For Primary 1 detailed sampling use the present year’s P1 figures. Exclude- composite not in year group private schools- excluded as large % pupil’s addresses and dentists outwith local area where school is. There is an argument if large % are day pupils in private school then need to consider if just sampling from state schools is a truly representative sample of the area’s population. Does anyone inspect private schools? Lothian’s problem. If do sample and do include how show that oral health improvement actual not as a result inclusion private schools? Disregard names of pupils and DOB columns Set up spreadsheets- separate worksheets per school Totalled number pupils in each class and number of classes in year

Slide 7 From individual pupil and DoB (alphabetical) in each school in NHS Board area to Spreadsheet- individual work sheets per school (number of pupils each class per school) Then amalgamate into one spread sheet with list of all schools, all classes, total year population and class size

Slide 8

Example

School + Class CHP or CPP Number in Year Number in Class A1 X 19 19 A2i Z 51 24 A2ii Z 51 27 A3 Y 31 31

For this example have labelled schools A through to Z and numbered schools according to number of schools per letter in alphabet. Classes have been labelled by roman numerals for this example and 3 CHPs or CPPs are called X, Y, Z.

Slide 9

Sampling- Step 2 Divide according to school year size

• Divide into:
• Small (1-15)
• Medium (16-49)
• Large (50+)

Slide 10 First 4 columns go from 1-217 (every class listed) according to class size. Then copied and pasted 3 blocks of columns, separated by yellow- F-I schools with year size 1-15, Columns K-N cover schools with year size 16-49 etc.

Slide 11

Summary school information

• No. pupils in school year

Total Small (1-15) Medium (16-49) Large (50+) Year Population 319 2396 1017 3732

• No. of

schools 40 84 17 141

Now have the information at Board Level broken down into size of year group (will be more than one class for medium and large categories).

Slide 12

Calculate no. of pupils and schools to inspect

Year Size

• No. of

schools

• No. of

children % of Board year population Sample required Small 40 319 8.5 26 Medium 84 2396 64.2 193 Large 17 1017 27.3 82 Total 141 3732 100 301

Now have the information at Board Level broken down into year group sizes (small, medium & large). The next step is to calculate the proportion of the total school population of that year group attending each of the school sizes. This is calculated by dividing the number of children in each school size by the total NHS Board year group population. Here I have multiplied this to get a percentage. Following the BASCD guidelines then looking to inspect a sample of 300 for NHS board

• area. Numbers will come out over the 300 as whole classes are included.

To then work out how many children in each

Slide 13

Sampling Step 3 Random numbers 3.1

• Generate random numbers (>0 to <1)
• Select column to input random no.
• Use Σ function (more functions)
• Select RAND
• Press OK

Using excel (at the NDIP Coordinators’ Study Day 2010 there was discussion around using other programmes/tables for random numbers other than just excel) For NHS Board Level assign random number for each of class in each size category (sort within category)

Slide 14

Sampling Step 3 Random numbers 3.2

• Will give random no. in top cell
• Select that cell
• Click on little box bottom right
• Click and drag down column
• Populate random no. in all cells

Slide 15

Sampling Step 3 Random numbers 3.3

• To set random number values

– Highlight random numbers – Copy column – Paste Special – Select Values option – Random numbers will be held

I used ‘print screen’ from the excel spreadsheets as the way I generated random numbers meant I had to fix their values otherwise they changed. Iain Hay, who attended the day showed me another way to generate random numbers in excel which I will put in a revised presentation for the NDIP manual.

Slide 16

Sampling Step 3 Random numbers 3.4

• Each school class assigned random

number

• Sort by random no. column

Slide 17 The CHP column can be substituted for CPP level information

Slide 18

Calculate no. of pupils to inspect

• No. pupils in school year

Total Small (1-15) Medium (16-49) Large (50+) Year Population 319 2396 1017 3732 Proportion of year population (%) 8.5 64.2 27.3 100

• No. of children

required to sample 26 193 82 301

Slide 19 Step 4- read off from top the number required to fulfil your sample calculation

Slide 20

Calculate no. of pupils and schools to inspect

Year Size

• No. of

schools

• No. of

children % of Board year population Sample required

• No. selected

schools (pupils) Small 40 319 8.5 26 6 (28) Medium 84 2396 64.2 193 10 (199) Large 17 1017 27.3 82 4 (88) Total 141 3732 100 301 20 (315)

We can see that we have sampled slightly more than we calculated- whole class approach, but allows for absenteeism

Slide 21

Sampling at CHP/CPP Level

• Why?

– Local comparisons – Local planning

• How?

– Minimum of 250 per CHP/CPP area – Weighting (carried out by ISD) Why carry out stratified sampling? Get more info at smaller area Sub analyses- so need to inspect large numbers of children Weighting will be covered later on but required as CHP/CPPs are of unequal size

Slide 22

Sampling protocol

• State maintained schools
• Minimum sample of 250 per CHP/CPP
• Select additional 15-20%
• All children in class
• Refresh sample each year

Slide 23

Sampling- CHP/CPP worked example

• Lists from Local Education Dept.
• Exclusions
• 3 CHP/CPP areas
• 141 schools
• 3732 pupils
• 216 classes

Most areas want sub-analyses therefore need to inspect larger number of children- minimum 250 per CHP/CPP area so looking at calculating sampling 300 per area. 14 NHS Regional Boards and 40 CHP areas and 32 CPPs

Slide 24 Looked at worksheet titled schools that listed all schools in alphabetical order. For the first stage in sampling, schools are first divided into different CHPs/CPPs and then sampling proceeds for each CHP/CPP area separately as will be described. Sorted by CHP/CPP (column B) to separate worksheet into X Y and Z. Created separate worksheets for each CHP/CPP area

Slide 25

CHP/CPP Level

• Sort into separate CHPs/CPPs
• Sort by school year size
• Allocate random numbers
• Sort by random numbers
• Select classes to inspect

Once sorted by CHP/CHP can carry out similar sorting to NHS Board area example, so sort into year sizes Allocate random number per class, set random number value by copy and paste special, and select VALUE Sort by random number (Smallest to largest) Calculate sample size (300) to allow for absenteeism, declining and go through list adding class totals until number is reached.

Slide 26

CHP/CPP pupils in school year

CHP /CPP School Year Size Total Small Medium Large X 37 561 468 1066 Y 139 1024 419 1582 Z 143 811 130 1084 Total 319 2396 1017 3732

Slide 27

• No. pupils to inspect per class

size per CHP/CPP to get sample 300

CHP/CPP School year size Total Small Medium Large X 10 158 132 300 Y 26 194 80 300 Z 40 224 36 300

Slide 28

Slide 29

Completed sampling for CHPs/CPPs

• No. selected schools (pupils)

Year Size X Y Z Small 3 (11) 2 (27) 9 (44) Medium 10 (180) 11 (208) 12 (245) Large 7 (151) 3 (84) 2 (51) Total 20 (342) 16 (319) 23 (340)

Slide 30

Weighting

• Results need to be weighted for average for

NHS Board area

CHP/CPP Sample size School Yr % inspected X 300 1066 28 Y 300 1582 19 Z 300 1084 28 Total for NHS Board 900 3732 24

3 CHP/CPP areas used in this example are unequal in size and might be variations in demographics within CHP/CPP areas so the dmft/DMFT means can not simply be added together and divided by 3 to obtain an overall NHS Board average as there will be over representation of areas with smaller populations. Therefore the contribution of each CHP area must be weighted according to the size of the population in relation to the whole NHS Board population.

Slide 31

Reference

• Pine CM, Pitts NB, Nugent ZJ.

BASCD Guidance on Sampling for Surveys of Child Dental Health. Community Dental Health 1997; 14 (Supplement 1):10-17. Found www.bascd.org in policy section under guidance dated Friday 23rd January 2004.

Slide 32

Useful websites

• www.scottishdental.org
• www.dundee.ac.uk/dhsru
• www.bascd.org
• www.glac.ac.uk/departments/dentalsc

hool

Slide 33

Appendix 1- CHP/CPPs

Slide 34

Slide 35