DEATHS - 7 th Listing (6 th Update) & CANCER 4 th Listing (3 rd Update) JUNE 2009

Transcription

1 UK Data Archive Study Number Health and Lifestyle Survey Deaths and Cancer Data, June 2009 DEATHS - 7 th Listing (6 th Update) & CANCER 4 th Listing (3 rd Update) JUNE 2009 WORKING MANUAL THIS MANUAL AND THE DATA FILE ARE DESIGNED TO BE USED IN CONJUNCTION WITH THE HALS1 & HALS2 MANUALS AND DATABASES The Health & Lifestyle Survey University of Cambridge Wolfson College Barton Road Cambridge CB2 2SR bdc10@ cam.ac.uk Hals Deaths7 & Cancer

2 Acknowledgments The Health and Lifestyle Survey team wish to express their thanks to the staff at ONS (OPCS) Southport whose efforts at tracing and flagging the HALS respondents have made possible the establishment of the Deaths and CancerData File List of Contents Acknowledgements 2 Introduction. 3 Coding.4 Cancer Registration... 4 Current File Status 4 Data File Format 5 Causes of Death 5 Updating the Dataset 5 Data File Columns & Variables 6 SPSS System File Information 9 Hals Deaths7 & Cancer

3 HEALTH AND LIFESTYLE (HALS) DEATHS AND CANCER DATA FILES SEVENTH DEATHS REVISION & FOURTH CANCER REVISION JUNE 2009 Introduction A high proportion of the original 9003 respondents in the 1984/85 Health and Lifestyle Survey have been "flagged" on the NHS Central Register at ONS (formerly OPCS) Southport. This now allows analysis of final outcome - death - to be correlated against previously reported medical history, physiological status and lifestyle behaviour. To date some 97.8% of the sample has been flagged. Flagging of the sample took several years because at the outset no NHS numbers were collected for the respondents, as it was not envisaged at that time that follow-ups, or flagging, would be possible. The, then, current address and name were recorded on a separate data file which was subsequently used, together with the information on date of birth, to identify the respondents on the NHS Central Register. Unfortunately, the names by which the respondent was usually known, as given to the interviewer when the first contact was made, was not necessarily the names by which the respondents were registered on the NHS data file, e g. John Brown might have been registered as Aloysius John Brown. Not only that, the address may not have related to where the person was registered at the, then, local FHSA. Thus the process of flagging took some time and resulted in many checks in order to ascertain that the flagging registration indeed related to the respondent seen at the Survey. The computerisation of the current NHS Central Register in 1991 also added to the problems. The process of manually identifying the respondents on the Register slowed down considerably during the computerisation process when the OPCS staff were heavily involved in that major exercise. However, this has meant that subsequently the outstanding non-flagged respondents could be more readily identified and we hope that the file is nearly complete in terms of flagging. Some respondents, at the outset, refused to give their names, some also only gave a Christian name, and consequently there are about 40 cases which can never be flagged on the register. Therefore they have to be removed from the denominator used as the reference point for calculating survival etc of the HALS population. In 1991/2 a re-run of the Survey was made (HALS2) which resulted in 5352 respondents being re-interviewed. During the fieldwork the interviewers attempted to trace those respondents for whom we did not have a current address. In some cases where the respondent was reported as being dead, it has been possible to verify this statement with the ONS Register. However, in a few cases, as Mark Twain said, "reports of my death have been greatly exaggerated" so, although the respondent was reported dead by neighbours at fieldwork, they may in fact still be alive. Hals Deaths7 & Cancer

4 Coding The files are constantly being updated as we are notified of cancer registrations and deaths by ONS. On the facsimile death certificates and cancer notifications that we receive are indicated the cause of death codes, and the cancer type as identified by ONS. Initially the notifications were made under the Ninth ICD Revision and Classification. By 2005 all notifications have been under the Tenth ICD Revision and Classification. However, in order to allow for ease of analysis and continuity, all deaths have been recoded to ICD9. Cancers are listed under both ICD9 and ICD10 codes. Up to the beginning of July 2009 we have received notification of some 2883 deaths, and 1468 respondents have been coded for cancer. It is at present unlikely that we will undertake the entire re-coding of the deaths from ICD9 to ICD10, although for many cases we have the initial coding in the ICD10 format. Cancer Registration The returns from the various Cancer Registries to ONS have in the last few years been more reliable and the backlog of earlier notifications has been mostly cleared. This is the third file update to include a Cancer Data File for HALS. As stated previously, the regional input varies, and thus analyses by Region are still not recommended. Once again there is a delay in notifying us of cancer registrations, although the delay seems to be reducing. Whilst those up to the end of 2006 seem to be as complete as we are ever likely to get them, For 2007 there has been 50 notifications so far, and 7 for In 2009 there are none at all, whereas notification of deaths are within a few months of being up-to-date. For 2007 we have been probably notified of all the deaths. For cancer registrations there are always going to be missing cases due to the patchy returns from some regional Registries, and these factors need to be borne in mind when analyses using the cancer data are undertaken. However, the compilation of the cancer returns is showing that many respondents who were registered as having cancer have died, without any indication of cancer on the death certificate, even as a notified associated condition that did not contribute to the death. Providing that the death certificate is accurate, the encouraging implications are that many cancers are being successfully treated and the respondent has died from other non-cancer related causes. To enable studies to be carried out using all the information provided under the two coding formats has meant that we have had to convert the earlier ICD9 to the equivalent ICD10 coding, and also to recode the later ICD10 notifications to ICD9. Thus the cancer information is available on this data file in both formats. We have endeavoured to accurately identify the appropriate equivalent coding, however there are some minor anomalies because of the way the information has been provided to us. There will be found some minor discrepancies in relation to codings for cancer just prior to death and the codings appearing on the death certificates (Cause of death registered by a doctor and cancer type identified by a pathologist). The age of the respondent to one decimal place is provided at registration together with the month and year of registration. Thus on the file in this format is the age at HALS1 (1984/5), the age at each cancer registration (up to three separate registrations are recorded in some cases), and the age at death. The age at HALS2 (1991/2) will be exactly seven years older than at HALS1 because the respondents at HALS2 were seen at the same time of the year as at HALS1. Of course formal registration of cancer does not indicate the time of the onset of the condition which is inevitably unknown. Hals Deaths7 & Cancer

5 Current File Status On the data file is a code (variable FLAGCODE) which enables the current status of the respondent on the HALS files to be identified : Code 1 (ON FILE) - Currently alive and flagged on the NHS Register 2 (NOT NHS REGIST) Not currently registered with the NHS but not known to be dead 3 (DECEASED) Known dead and death certificate information recorded on file 4 (REP DEAD NOT ID) Reported dead to HALS not on NHS Register (may be alive) 5 (EMBARKED-ABROAD) Identified on NHS Register but currently out of country 9 (NOT YET FLAGGED) Not currently flagged for various reasons (no name etc.) Data File Format The data file is organised as in the format described in the following pages, with all 9003 cases originally listed as a flat file 152 columns wide and with the HALS serial number in the first five columns to enable linking to be made to the other HALS data files (ESRC nos and 3279). There is a basic data list file provided which will identify the location on the data file of the specific variable together with the variable and value labels. This will enable the data to be complied as an SPSS system file. An already prepared SPSS system file of the data is also provided. Causes of Death All causes listed on the Death Certificate and coded by ONS are entered on the data file. The cause selected as "Underlying" has been derived by ONS who in 1993 introduced an automatic coding system developed by the US National Center for Health Statistics called the Automated Cause Coding System(ACCS). The main effect has been to the interpretation of WHO Rule 3 which has meant that, for example, in 1993 the death rate from pneumonia (based on Underlying Cause) was twice that in It is therefore recommended that users of the data file do not entirely rely on the listed Underlying Cause for analysis purposes but check on all causes that are entered and the order which they are listed (as 1a, 1b, 1c, 2a, etc.). All categories can also have joint causes and so there are two fields for each category of cause (e.g. Cause 1a and Joint Cause 1a). The initial cause will be of five columns width to allow for "E" codes (e.g. E9889), the joint cause will only be four columns wide (e.g. 1990). In all cases there is an implied decimal after the third numeral as the fourth column relates to specific attributes of the condition. Because of the problems of the "E" codes and because some codes start with a zero (e g. 038), and some end with a zero, (which is different from those which are only three columns wide e.g. 480 and 480.0), the causes are most easily handled as alphanumeric variables. Updating the Dataset We are not sure, if and when, the data file will be updated again. But cancer and death registrations are continuing to be received at present from Southport (now The NHS Information Centre ) and this current version - dated June is the fourth to list cancer registrations and the seventh to update the deaths. In 2008 the NHSCR was re-organized as The NHS Information Centre. Hopefully we will continue to receive the deaths and cancer returns. Since the Survey was first registered, the center at Southport has undergone several name changes from OPCS to ONS and now to NHSIC. Hals Deaths7 & Cancer

6 Data File Columns and Variables Column Variable Range and no name missing value 1-5 SERNO HALS Serial number (Date of birth These dates may not be the same as originally registered In the HALS 1 file (1984/5). but by subsequent reference to various information sources are those adjudged as being the most likely true dates of birth) 6-7 BIRTHDAY Data removed for Data Protection Purposes 8-9 BIRTHMTH 1-12 Calendar month of Birth BIRTHYR Year of Birth Date of Death DEATHDAY Data removed for Data Protection Purposes DEATHMTH 1-12 Calendar month of Death DEATHYR Year of Death HALS1AGE Assessed age at HALS1 in years and tenths of a year 97.8 (one decimal place is implied) (Obtained by subtracting date of Birth from date seen at HALS1. Differs from age listed on the HALS1 file (whole reported years) to enable survival analyses to be made. Also verification of the date of birth from the ONS Register has revealed that in some cases the reported age by the Respondent was incorrect. Thus age from HALS1AGE is the more accurate value) DEATHAGE 0.0, Age at Death in years and tenths of a year (one decimal place is implied) (Obtained by subtracting date of Birth from date of Death - Code "0.0" not known to be dead) Causes of Death Listed under the 9 th ICD system of coding in the order indicated by ONS. Underlying Cause initially, with all other Causes listed in the categories as on the Death Certificate. Where joint causes are listed those with "E" coding are listed under the first variable for that category of cause. These variables are handled as alphanumerics for analysis purposes 28-32(A) UCAUSE Underlying Cause of Death E (A) JUCAUSE Joint Underlying Cause of Death (A) CAUSEIA Cause la E (A) JCAUSE1A Joint Cause la (A) CAUSEIB Cause lb E (A) JCAUSE1B Joint Cause lb (A) CAUSE1C Cause lc E9888 Hals Deaths7 & Cancer

7 Column Variable Range and no name missing value 60-63(A) JCAUSE1C Joint Cause lc (A) CAUSE2A Cause 2a E (A) JCAUSE2A Joint Cause 2a (A) CAUSE2B Cause 2b E (A) JCAUSE2B Joint Cause 2b FLAGCODE 1-9 Flagging status of respondent derived from ONS Coding & Value Labels 1 ON FILE Currently alive and flagged or being flagged 2 NOT NHS REGIST Not registered with NHS - not known to be dead 3 DECEASED Known dead 4 REP DEAD NOT ID Reported dead but not identified on register (may be alive) 5 EMBARKED-ABROAD Identified on NHS Register but currently out of country 9 NOT YET FLAGGED Not currently flagged for whatever reason C1AGE 0, Age at 1 st cancer registration in years and tenths of a year (A) ICD st Cancer registration ICD9 coding (A) ICD101 C003-1 st Cancer registration ICD10 coding Q C2AGE 0, Age at 2 nd cancer registration in years and tenths of a year (A) ICD nd Cancer registration ICD9 coding (A) ICD102 C030-2 nd Cancer registration ICD10 coding D C3AGE 0, Age at 3 rd cancer registration in years and tenths of a year (A) ICD rd Cancer registration ICD9 coding (A) ICD103 C140-3 rd Cancer registration ICD10 coding D374 Hals Deaths7 & Cancer

8 Column Variable Range and no name missing value REG1MTH 1-12 Month of 1st registration REG1YR Year of 1st registration REG2MTH 1-12 Month of 2nd registration REG2YR Year of 2nd registration REG3MTH 1-12 Month of 3rd registration REG3YR Year of 3rd registration SEX 1-2 Sex of respondent Coding & Value Labels 1 MALE 2 FEMALE FLAGCODE - current flagging status June 08 Value Label Value Frequency Percent Percent on file not NHS regist deceased rep.dead not id embarked -abroad no flag yet rec Total FLAGCODE - current flagging status June 09 of those registered for cancer Value Label Value Frequency Percent Percent on file (alive) not NHS registered deceased Total Hals Deaths7 & Cancer

9 HALS2009.SAV'. -- SYSTEM FILE INFORMATION SYSFILE INFO -- File c:\halsded9\dead09.sav Label: HALS 7th Death and 4th Cancer update to JUNE 2009 Created: 06 Oct variables and 9,003 cases File Type: SPSS Data File N of Cases: 9003 Total # of Defined Variable Elements: 36 Variable Information: Name Position SERNO * No label * 1 Format: F5 BIRTHMTH month of birth 2 BIRTHYR year of birth 3 DEATHMTH month of death 4 DEATHYR year of death 5 HALS1AGE age at HALS1 6 Format: F5.1 DEATHAGE age at death 7 Format: F5.1 UCAUSE underlying cause 8 JUCAUSE joint underlying cause 9 CAUSE1A cause 1a 10 JCAUSE1A joint cause 1a 11 CAUSE1B cause 1b 12 JCAUSE1B joint cause 1b 13 CAUSE1C cause 1c 14 JCAUSE1C joint cause 1c 15 Hals Deaths7 & Cancer

10 HALS2009.SAV - SYSTEM FILE INFORMATION - continued CAUSE2A cause 2a 16 JCAUSE2A joint cause 2a 17 CAUSE2B cause 2b 18 JCAUSE2B joint cause 2b 19 FLAGCODE current flagging status Jul Format: F1 Value Label 1 on file 2 not NHS regist. 3 deceased 4 rep.dead not id 5 embarked -abroad 9 no flag yet rec. C1AGE 1st reg cancer age 21 Format: F5.1 ICD91 1st ICD9 code 22 ICD101 1st ICD10 code 23 C2AGE 2nd reg cancer age 24 Format: F5.1 ICD92 2nd ICD9 code 25 ICD102 2nd ICD10 code 26 C3AGE 3rd reg cancer age 27 Format: F5.1 ICD93 3rd ICD9 code 28 ICD103 3rd ICD10 code 29 REG1MTH * No label * 30 REG1YR * No label * 31 Format: F4 REG2MTH * No label * 32 REG2YR * No label * 33 Format: F4 Hals Deaths7 & Cancer

11 HALS2009.SAV - SYSTEM FILE INFORMATION - continued REG3MTH * No label * 34 REG3YR * No label * 35 Format: F4 SEX sex of case 36 Format: F1 Value Label 1 male 2 female Hals Deaths7 & Cancer