Description and quality assessment of MORGAM data

Record inventory, routine data checking and changes made to the data Zygimantas Cepaitis1, Bijoy Joseph1,2, Matti Niemelä1, Olli Saarela1 and Kari Kuulasmaa1 for the MORGAM Project3

¹ Department of Chronic Disease Prevention, National Institute for Health and Welfare (formerly Department of Health Promotion and Chronic Disease Prevention, National Public Health Institute), Helsinki, Finland
² Since January 2007 at Indic Society for Education and Development (INSEED), Nashik, India
³ See Annex for the sites and key personnel of contributing MORGAM Centres

© National Institute for Health and Welfare and the MORGAM Project investigators Last updated: 24 February 2009 For more information, please contact Zygimantas Cepaitis (firstname.lastname@thl.fi) or Kari Kuulasmaa (firstname.lastname@thl.fi)

Contents

• 1. Introduction
• 2. Record inventory
  • 2.1 Comparison of other forms with Form 20
  • 2.2 Comparison of event data with event inventory data
  • 2.3 Comparison of mortality data and fatal event records
• 3. Routine checking of the correctness of the data
• 4. Removal of records from the MORGAM database and changes made to the data
• Updates to this document
• Specific tables
  • Table I.1 Serial number inventory: comparison of other forms with Form 20
  • Table I.2 Serial number inventory: comparison of event data with event inventory data
  • Table I.3 Comparison of deaths in form 25 and records of nonfatal events within 28 days
  • Table I.4 Number of resolved and unresolved constraint violations
  • Table I.5 Removal of records from the MORGAM database and changes to the data after they have been used for analyses

1. Introduction

The process of transfer of data from the MPCs to the MDC involves routine checking of

• the completeness of the data to see that all necessary data forms have been transferred for every member of the cohorts, and
• the correctness of the data, to see that there are no undefined or unrealistic data values or that there are no logical inconsistencies between the data items.

We refer to the first of these with term "record inventory". Routine data checking is part of the portable MORGAM data transfer system which is available to the MPCs. Therefore, the MPCs can correct most of the inconsistencies in the data already before sending the data to the MDC. The remaining inconsistencies are usually resolved through communication between the MPCs and the MDC. The current status of the the record inventory and routine data checking is described here.

2. Record inventory

2.1 Comparison of other forms with Form 20

A reference for the MORGAM Record inventory is the MONICA survey data (Form 20), which includes very basic information of all members of the MORGAM Cohorts. The MORGAM cohorts are assumed to consist of the subjects on whom Form 20 data are available. Other data that should be available on every member of the Cohorts are defined in data transfer formats for:

• Additional baseline data (Form 21),
• Follow-up data (Form 25),
• Coronary events inventory (Form 27) and
• Stroke events inventory (Form 28),

and optionally in the data transfer formats for:

• Baseline Data Addition 2 (Form 30)
• Venous Thromboembolic Events Inventory (Form 29)

If these these optional data are provided, they should be provided for every member of the cohorts concerned.

The matching of the records of the above data forms with Form 20 is shown in Table I.1. Ideally, the columns corresponding to Extra and Missing should have zero entries, except for the cohorts where the optional thromboembolic events were not followed up. There are currently no inconsistencies.

2.2 Comparison of event data with event inventory data

Coronary event data records (Form 22), Stroke event data records (Form 23) and Venous thromboembolic event records (Form 26) should be available only for persons that had such events during the follow-up. On the other hand, one person may have several event records. The number of event records of each kind, that should be available for each person, is specified in the event inventory data, i.e. in items COREV of Form 27, STROKES of Form 28 and VTEV of Form 29.

The matching of the event records with the event inventory data is shown in Table I.2. Ideally, the columns corresponding to Extra and Missing should have zero entries, except for the cohorts where the optional thromboembolic events were not followed up. There are currently no inconsistencies.

2.3 Comparison of mortality data and fatal event records

According to the MORGAM instructions:

• A coronary event record should be provided for every death during the follow-up period which (see item COREV of Form 27)
  • has a CHD code in the death certificate or final death diagnoses. This includes ICD-8 and 9 codes 410-414 and ICD-10 codes I20-I25, or
  • has a code of sudden or unattended death or cardiac arrest as an underlying cause of death in the death certificate or final death diagnoses. This includes ICD-8 code 795, ICD-9 code 798 and ICD-10 codes I46, R96, R98 and R99.
• A stroke event record should be provided for every death which has a code for stroke in the death certificate or final death diagnoses. This includes ICD-8 codes 430-436 and ICD-9 codes 430, 431, 433, 434 or 436 and ICD-10 codes I60, I61, I63 and I64 (see STROKES of Form 28).
• A venous thromboembolic event record should be provided for every death which has a code for pulmonary embolism or deep vein thrombosis in the death certificate or final death diagnoses. The relevant ICD codes are
  • ICD-8: 450, 451, 671 and 673.9
  • ICD-9: 415, 451 (excl. 451.0), 671.3, 671.4 and 673.2
  • ICD-10: I26, I80 (excl. I80.0), O87.1 and O88.2.

The collection of all these event records even if the validation of the diagnosis has revealed that ICD code indicated a false positive diagnosis facilitates this inventory and hence the confirmation that suspect coronary, stroke and thromboembolic deaths have been validated systematically.

The deaths and their ICD codes are available in the Follow-up data (Form 25). From some MPCs only the underlying cause of death is available, whereas from many also the other causes specified in Form 25 have been reported to the MDC. In checking that event data have been provided for the relevant causes of death, we use all available causes of death except the contributing cause of death (DEATHDO).

The event records, including the fatality of the event, are specified in Form 22, Form 23 and Form 26. The event corresponding to the death must be fatal and the date of onset must be less than 28 days and not later than the date of death.

On the other hand, for every event record indicating a fatal event, there must be a death in less than 28 days in Form 25.

The consistency between the Follow-up data (Form 25) and the event records of fatal events ( Form 22, Form 23 and Form 26) is shown in Table I.3. Ideally, the columns corresponding to Missing should have zero entries, except for the cohorts where the optional thromboembolic events were not followed up. In these cohorts the column Total of fatal VTE events in Form 26 should be zero. There are minor inconsistencies:

• for coronary events in FRA-LIL, FRA-STR, ITA-FRI, ITA-PAM, ITA-ROM, SWE-NSW and UNK-BEL;
• for stroke events in FRA-STR, ITA-BRI, ITA-FRI, ITA-PAM, ITA-ROM, UNK-GLA and UNK-SHH; and
• for thromboembolic events no inconsistencies (Note that the comparison is relevant only for the few cohorts that had the optional follow-up for thromboembolic events.)

3. Routine checking of the correctness of the data

The event data were checked routinely for the constraints specified in the MORGAM Manual. All violations of the constraints were checked for their correction or elucidation by MPCs and were reported to the MDC. For many data items, values outside the constraint limits were acceptable, but the MPC had to check and confirm that the values were not unusual due to data errors. The MPCs were asked to revise the data only if they were incorrect.

The numbers of current resolved and unresolved constraint violations are given in Table I.4.

4. Removal of records from the MORGAM database and changes made to the data

Removal of records from the MORGAM analysis data sets that may have been used for earlier publications are described in Table I.5. Such records have been flagged as "removed" in the MORGAM database. Table I.5 also shows significant changes that have been made to the data since they have already been used for some publications.

Updates to this document