Relying Solely on Postal Code to Assign People/Events to Census - - PowerPoint PPT Presentation

▶
relying solely on postal code to assign people events to
SMART_READER_LITE
LIVE PREVIEW

Relying Solely on Postal Code to Assign People/Events to Census - - PowerPoint PPT Presentation

Jun 28, 2023 335 likes •652 views

Relying Solely on Postal Code to Assign People/Events to Census Geography or Ministry Geography Health Analytics Branch Health System Information Management & Investment Division Ministry of Health and Long-Term Care 2011 APHEO Conference

slide-1
SLIDE 1

Relying Solely on Postal Code to Assign People/Events to Census Geography or Ministry Geography

Health Analytics Branch Health System Information Management & Investment Division Ministry of Health and Long-Term Care 2011 APHEO Conference May 15-16, 2011

slide-2
SLIDE 2

What will we cover?

  • Why do we need crosswalks between different

geographies?

  • How do we assign people to different geographies based
  • n their postal code?
  • Limitations and considerations when using the PCCF and

PCCF+ for assigning people and events

  • What does this mean for public health epidemiologists

when assigning people and events to small areas?

2

slide-3
SLIDE 3
  • Different data sources collect different geographic locators:
  • Census data are available at, among others, CD, CSD and DA levels
  • Vital statistics until 2007 included municipality and postal code but from 2008 will have
  • nly postal code
  • Utilization data varies:
  • Hospital separations (DAD) and emergency room visits (NACRS) - ministry residence

code (municipality) and postal code

  • Medical claims (OHIP) and Continuing Care Reporting System (CCRS) - postal code
  • Population health data - postal codes and/or municipalities
  • Geographic level at which data are available may not be that

needed for analysis

  • Different data sources may have to be linked to paint a

complete picture to answer research questions or to support program planning/decision making

  • Postal Codes → census geography and ministry geography

Why do we need crosswalks between different geographies?

3

slide-4
SLIDE 4

How do we assign people to different geographies based on their postal code?

  • PCCF and Ministry Conversion Files – using Single Link Indicator (SLI)
  • “The Postal Code Conversion File (PCCF) is a digital file which provides a correspondence between the

Canada Post Corporation (CPC) six-character postal code and Statistics Canada’s standard geographic areas (e.g. census subdivisions, dissemination areas, dissemination block and census tract) for which census data and other statistics are produced. Through the link between postal codes and standard geographic areas, the PCCF permits the integration of data from various sources.”

  • Extracted from Postal Code Conversion File (PCCF) Reference Guide (July 2009 Postal codes)
  • MOHLTC creates, on an annual basis, a series of conversion files that link postal code to ministry

geographies: residence codes (municipality), public health unit and LHIN.

  • PCCF+ - using random allocation
  • “PCCF+ Version 5 consists of a SAS control program and a series of reference files derived from the

most recent Statistics Canada Postal Code Conversion File (PCCF) and 2006 postal code population weight file (WCF).” - Extracted from PCCF + Version 5F User’s Guide (July 2009 Postal Codes)

  • GIS – geographic location
  • The latitude and longitude of the postal code, available in the PCCF, allow postal codes to be mapped
  • spatially. With ArcGIS, or other GIS software, postal codes can be linked to other geographies based
  • n their geographic location.

4

slide-5
SLIDE 5

One CSD to Residence code, LHIN, PHU (MOHLTC) DA DA DA to LHIN Correspondence File (Stats Can) Postal Code to LHIN (MOHLTC) Postal Code CD/CSD/DA LHIN Residence Codes PHU (MOHLTC) Postal Code All CD/CSDs All Residence Codes LHIN PHU (MOHLTC) MOHLTC Residence Codes (MOHLTC) 2006 Census CSD (Stat Can)

How are the Ministry Conversion Files Created?

5

PCCF (Stat Can) SLI Postal Code, CSD, DA (MOHLTC) Postal Code to Residence Code & PHU (MOHLTC) CSD CSD

slide-6
SLIDE 6

PC (Canada Post) CD (StatsCan) CSD (StatsCan) CT (StatsCan) DA (StatsCan) Residence Code (MOHLTC) PHU (MOHLTC) LHIN (MOHLTC) PCCF/PCCF+ (StatsCan) Conversion Files (MOHLTC)

Update/Release Cycles for Different Geographies and Related Products

2006 2010 2007 2008 2011 2009 2012

6

slide-7
SLIDE 7

What should we be aware of when using PCCF and PCCF+ to assign people to census geography and MOHLTC geography?

  • Objective: To demonstrate the advantages and limitations of different

methods available for assigning people/events to other geographies based solely on postal code

  • Methods
  • To assess completeness of coverage and consistency of assignment in the same

year , PCCF 2008, PCCF 2009 and PCCF 2010, and their related ministry conversion files were analyzed.

  • To assess stability of assignment over time, longitudinal analyses were

conducted.

  • For longitudinal comparison, PCCF 2009, was used as the reference.

Assignment based on PCCF 2009 was compared with that based on PCCF 2008 and PCCF 2010 .

  • Postal codes can be retired and reborn. PCCF includes retired as well as active
  • records. Only active records are included in analysis conducted for this review.

7

slide-8
SLIDE 8

Limitations of Postal Code Conversion Files (PCCF/PCCF+)

“PCCF contains multiple records for a postal code when the postal code straddles more than one block-face, dissemination block, or dissemination area”.

  • Extracted from Postal Code Conversion File (PCCF) Reference Guide (July 2009 Postal codes)
  • Over a third of all postal codes have multiple records
  • Especially postal codes that cover multiple address ranges, rural

routes and community mailboxes

  • For example: Delivery via a Rural Route – e.g. K8A6W3

PCCF 2009: 39 Active Records

8

slide-9
SLIDE 9

Delivery Model Type (DMT) for Active Postal Code (PCCF 2009)

9

Number of Active Postal Codes Delivery Model Type Description PC with Multiple Records All Active PC Percent Delivery to Face-Block Address 85,849 250,620 34.3% Delivery to an Apartment Building 753 8,575 8.8% Delivery to a Business Building 605 5,146 11.8% Delivery to a Large Volume Receiver 606 3,498 17.3% Delivery via a Rural Route 464 472 98.3% General Delivery 104 156 66.7% Delivery to a Post Office Box (not a Community Mail Box) 1,644 2,572 63.9% Delivery to a Large Volume Receiver (post office box) 2,036 2,919 69.7% Delivery via a Suburban Service 29 34 85.3% Rural Postal Code (second digit of postal code is “0”) 1,078 1,125 95.8% Total 93,168 275,117 33.9%

slide-10
SLIDE 10

Delivery via a Rural Route – e.g. K8A6W3 Address Ranges

Number Delivery Mode Delivery Installation Street Name City 2621-3859 Odd RR 2 BEACHBURG RD PEMBROKE 2654-3558 Even RR 2 BEACHBURG RD PEMBROKE 400 RR 2 COON HOLLOW TRAIL PEMBROKE 131-197 Odd RR 2 FINCHLEY RD PEMBROKE 186 RR 2 FINCHLEY RD PEMBROKE 1090-1889 RR 2 GREENWOOD RD PEMBROKE 20771-21055 Odd RR 2 HIGHWAY 17 PEMBROKE 89-476 RR 2 HILA RD PEMBROKE 1138-1838 Even RR 2 INDIAN RD PEMBROKE 1209-1837 Odd RR 2 INDIAN RD PEMBROKE 111 RR 2 INDUSTRIAL PARK RD PEMBROKE 361 RR 2 JEBWOOD TRAIL PEMBROKE 610-1292 Even RR 2 MCGONEGAL RD PEMBROKE 3033-3944 RR 2 MCLAUGHLIN RD PEMBROKE 27-170 RR 2 MEATH HILL RD PEMBROKE 499-850 RR 2 STURGEON MOUNTAIN RD PEMBROKE 1018-1071 RR 2 SUTHERLAND RD PEMBROKE 1002-1459 RR 2 VALLEY VIEW PEMBROKE 2252-2700 RR 2 ZION LINE PEMBROKE RR 2 STN MAIN PEMBROKE 10 10

Extracted from Canada Post website

slide-11
SLIDE 11

Delivery via a Rural Route – e.g. K8A6W3 Address Ranges

11 11

slide-12
SLIDE 12

How do PCCF and PCCF+ deal with multiple records?

  • PCCF and Ministry Conversion Files

“The single link indicator (SLI) was created to deal with postal codes with multiple records. The single link indicator identifies the geographic area with the majority of dwellings assigned to a particular postal code. Users should be aware that only a partial correspondence between the postal code and other geographic areas is achieved when using the single link indicator.”

  • Extracted from “Postal Code Conversion File (PCCF) Reference Guide” (July 2009 Postal codes)
  • PCCF+

“Records for most postal codes which serve more than one dissemination area – including most rural postal codes and several classes of urban postal codes – are assigned geographic codes based on a population-weighted random allocation among the possible dissemination areas and blocks.”

  • Extracted from “PCCF + Version 5F User’s Guide” (July 2009 Postal Codes)

12 12

slide-13
SLIDE 13

Are all units in each geographic level covered with the different assignment methods?

  • “Completeness of coverage”
  • Even when all records are included, PCCF does not cover

all units at every geographic level.

  • When the record with the SLI is used, more areas are

missed at every geographic level

  • With all 3 PCCFs examined, only partial coverage is

achieved at CSD, CT, DA and Residence Code levels. For example with PCCF 2009:

  • No individual or event will be assigned to 4 CTs even if all records are used; 20

CTs will be missed if only SLI records are used.

  • 46 CSDs are not covered if all records are used and 151 CSDs will be missed when
  • nly SLI records are used

13 13

slide-14
SLIDE 14

The table below shows the number of units covered by each PCCF by geographic level. Geographic Level # of Units PCCF 2008 PCCF 2009 PCCF 2010 All Records SLI Records All Records SLI Records All Records SLI Records CD 49 49 49 49 49 49 49 CSD 585 546 435 540 429 539 434 CT 2,136 2,132 2,117 2,132 2,116 2,131 2,117 DA 19,177 18,988 16,340 18,981 16,352 18,979 16,377 Residence Code* 684

  • 442
  • 440
  • 442

PHU* 36

  • 36
  • 36
  • 36

LHIN* 14

  • 14
  • 14
  • 14

Are all units in each geographic level covered with the different assignment methods?

14 14

Coverage is not an issue at PHU level (all PHU have postal codes linked to them).

* PCCF combined with Ministry conversion files

slide-15
SLIDE 15

The table below shows the number of units not covered by each PCCF by geographic level.

Are all units in each geographic level covered with the different assignment methods?

15 15

Geographic Level # of Units PCCF 2008 PCCF 2009 PCCF 2010 All Records SLI Records All Records SLI Records All Records SLI Records CD 49 CSD 585 39 150 45 156 46 151 CT 2,136 4 19 4 20 5 19 DA 19,177 189 2,777 196 2,825 198 2,800 Residence Code* 684

  • 242
  • 244
  • 242

PHU* 36

  • LHIN*

14

  • Coverage is not an issue at PHU level (all PHU have postal codes linked to them).

* PCCF combined with Ministry conversion files

slide-16
SLIDE 16

CSDs not Covered by All Records of Active Postal Codes in PCCF 2009

16 16

45 CSDs are not covered by records of active postal codes in PCCF 2009. * Not all are shown because of the scale of the map.

Indian Reserves & Settlements: 40 Unorganized: 3 Townships: 2 Total CSD: 45 Total population: 8,624

Even with PCCF+, coverage is an issue when we try to drill below PHU level.

slide-17
SLIDE 17

CSDs not Covered by SLI Records of Active Postal Codes in PCCF 2009

17 17

156 CSDs are not covered by the SLI records of active postal codes in PCCF 2009. * Not all are shown because of the scale of the map.

Indian Reserves & Settlements: 98 Unorganized: 7 Municipality: 1 Towns: 10 Townships: 38 Villages: 2 Total CSD: 156 Total population: 91,445

Coverage is even more of an issue with SLI when we try to drill below PHU level .

slide-18
SLIDE 18

Many postal codes are linked to multiple geographic units within the same PCCF

  • “Consistency of assignment in the same year”
  • In PCCF 2009, of the 275,117 active postal codes, 1,268

are linked to more than one CSDs.

  • With SLI in PCCF, all people/events from these postal

codes are linked to one CSD.

  • With PCCF+, people/events are distributed across

geographic units. As a result, two people living in the same household may end up being assigned to different units.

18 18

Consistency is not an issue at PHU level (all postal codes are linked to one PHU).

slide-19
SLIDE 19

Example of Multiple Active Records for a Single Postal Code – N0H2T0

19 19

This map shows the northern portion of Grey Bruce Health Unit. In PCCF 2009, N0H2T0 has 561 active records with which it is linked to 9 CSDs and 33 DAs. The 9 CSDs are shown on the map.

slide-20
SLIDE 20

Example of Multiple Active Records for a Single Postal Code – N0H2T0

20 20

The one active record with SLI = 1 link the postal code to a single CSD, CSD 3541055 (in brown).

slide-21
SLIDE 21
  • “Stability of assignment of over time”
  • Example: a person residing in any one of the 209 postal codes where the PCCF 2008 SLI

record points to one DA while the PCCF 2009 SLI record points to another, the individual would “appear” to have moved from one DA to another. Number of Postal Codes PCCF 2009 vs. PCCF 2008 PCCF 2009 vs. PCCF 2010 Same Unit Different Units Same Unit Different Units At CD Level 273,736 40 274,423 39 At CSD Level 273,688 88 274,353 109 At CT Level* 235,883 32 236,476 36 At DA Level 273,567† 209† 274,153 309 At Residence Code Level 273,688 88 274,353 109 At PHU Level 273,749 27 274,435 27 At LHIN Level 273,761 15 274,449 13

† Numbers of postal codes reported for by geography level are in relation to PCCF 2009.

* Not all postal codes are linked to a census tract.

How many postal codes are linked to different geographic units with PCCF for different years?

21 21

Stability is an issue at all levels of geography including PHU (27 postal codes are linked to different PHUs from PCCF 2008 to PCCF 2009).

* PCCF combined with Ministry conversion files

slide-22
SLIDE 22

Example of Postal Codes Linked to Different CSD in PCCF 2008 versus PCCF 2009 – K8A6W3, K8A6W4

22 22

This map shows part of the Renfrew County and District Health Unit. With PCCF 2009, K8A6W3 is linked to CSD 3547056 and K8A6W4 is linked to CSD 3547070.

slide-23
SLIDE 23

Example of Postal Codes Linked to Different CSD in PCCF 2008 versus PCCF 2009 – K8A6W3, K8A6W4

23 23

This map shows part of the Renfrew County and District Health Unit. With PCCF 2008, both postal codes, K8A6W3 and K8A6W4, are linked to CSD 3547075.

slide-24
SLIDE 24
  • Dealing with Invalid Postal Code

“Postal codes that do not match exactly to the PCCF or WCF, the first one to five characters of the postal code are used to try to assign partial geographic identifiers to the extent possible. This takes care of many situations where one or more characters of the postal code are invalid, but the first one to five characters are valid. Problem records include full diagnostic and reference information”. – Extracted from “PCCF + Version 5F User’s Guide”

  • User-Friendliness

PCCF, with SLI, may be used with any database/ statistical software. As well, using the SLI records from PCCF, MOHLTC creates postal code conversion files which serve as crosswalks between postal code and census geography and ministry geography. This files provide easy conversions between postal code and other geographies without the need for statistical software.

Other Considerations when choosing between PCCF, PCCF+ and Related Products

24 24

slide-25
SLIDE 25
  • To study the impact of Statistics Canada’s decision to remove CSD from all

vital statistics files (effective with 2008 births, stillbirths, deaths & marriages)

  • Method:

1. Compare the CSD assigned based on the postal code with the

  • 2. municipality code recorded as reference (still available in 2007 files):
  • 2. The postal code type that make up each of the four category was

analyzed.

Choosing between PCCF, PCCF+ and Related Products – Example: Vital Statistics, Deaths, 2007

25 25

PCCF Plus Match CSD Mismatch Total Match CSD 72,280 (83%) 2,119 (2%) 74,399 (86%) Mismatch 1,011 (1%) 11,533 (13%) 12,544 (14%) Total 73,291 (84%) 13,652 (16%) 86,943 (100%) PCCF (SLI)

slide-26
SLIDE 26

Error: No Match to PCCF, 73%

Warning: Non-residential, 6% Warning: Commercial or institutional, 6% Note: Multiple records, (population- based allocation), 10% Note: Multiple records, (population- based allocation), 13% Note: Multiple records, (population- based allocation), 87% Note: Multiple records, (population- based allocation), 83%

No error/warning message, 83% No error/warning message, 13% No error/warning message, 12% No error/warning message, 7% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Both match municipality recorded Both do not match SLI do not match PCCF+ match SLI match PCCF do not match Error: No Match to PCCF Error: Linked to PO geography Warning: Non-residential Warning: Business building Warning: Commercial or institutional Warning: Retired postal code Note: Multiple records (equal weight allocation) Note: Multiple records, (population-based allocation) No error/warning message

Choosing between PCCF, PCCF+ and Related Products – Example: Vital Statistics, Deaths, 2007

26 26

slide-27
SLIDE 27

MOHLTC Conversion Files (SLI) ArcGIS (spatial join) Residence Code LHIN PHU

Postal Code

Assigning Postal Codes to Census Geography and Ministry Geography

27 27

PCCF CD CSD CT DA PCCF+ SLI

slide-28
SLIDE 28

What does it mean for public health epidemiologists?

  • Which level of geography are you interested in?
  • When trying to assign people/events to “high level geographies”, e.g. CDs or

public health units, relying solely on postal codes is less of an issue.

  • Given the limitations of coverage and consistency we have demonstrated, serious

considerations have to be given when designing any data collection and maintenance initiatives – “Should we rely on postal code alone as the geographic locator?”

  • Since all methods available for assigning people/events, based solely on postal code,

to other geographies have their pros and cons, users should select the method most appropriate for their purpose.

  • Be aware of whether the geographic levels in the data set you are using were

collected directly or were derived from postal code. For example, medical claims, CCHS, vital stats. Find out what method was used to allocate people/events.

  • The choice depends on:
  • The quality of the data to be linked, specifically the quality of the postal code

field (PCCF+ can take care of invalid postal code).

  • Analytical capacity (PCCF does not require the use of SAS while PCCF+ does. With

GIS capabilities, postal code may be linked to other geographies based on their geographic location).

28 28

slide-29
SLIDE 29

For more information, please contact:

Elsa Ho

Health Analytics Branch Health System Information Management and Investment Division Ministry of Health and Long-Term Care Tel: 416-326-6542 E-mail: elsa.ho@ontario.ca

Carol Paul

Health Analytics Branch Health System Information Management and Investment Division Ministry of Health and Long-Term Care Tel.: 416-327-7733 E-mail: carol.paul@ontario.ca

29 29