October 28, 2019
- We fixed errors in VACFLUNUM, VACFLUSH12M, VACFLUSP12M, VACFLUMO, VACFLUSHMO, VACFLUSPMO and VACFLUMODE in 2012. Previously, roughly 7,000 cases that were not in universe for VACFLUNUM were erroneously assigned to "don't know" instead of "not in universe." As a result of the error, the frequency of "don't know" cases for VACFLUSH12M and VACFLUSP12M, which are dependent on the value of VACFLUNUM, were artificially high in 2012. These cases should have been assigned to "no" instead of "don't know." Additionally, about 7,000 cases were incorrectly assigned to "don't know" instead of "not in universe" for VACFLUMO as well as about 400 cases mistakenly assigned to "not ascertained" instead of "don't know." These errors impacted the frequencies of VACFLUSHMO and VACFLUSPMO. Cases for these two variables that should have been assigned to "don't know" were assigned to "not ascertained," making the number of "don't know" cases artificially low. Finally, the 58 "don't know" cases for VACFLUMODE were previously incorrectly assigned to the "refused" category. The updated frequencies for these variables now correctly reflect the underlying data.
September 3, 2019
- Newly released 2018 NHIS imputed income variables are now integrated into the IPUMS NHIS database. We have also added more than 100 variables from the 2018 cancer, asthma, immunization, and disability supplements.
- We fixed an error in the value label for values of 99 in AGE. Previously, we had identified 99 as "missing" when it is actually the top code for age in 1969-1995.
- We fixed an error where we imposed a top code of 11 or more to the number of PSA tests undergone in the past 5 years reported in PSANO5YR. Now values of PSANO5YR range from 0 to 95 in all years except for 2000, which is top coded in the underlying data at 11 or more tests.
- We fixed an error in which we had transposed the values for "Not mentioned" and "Mentioned" in 2015 for COLCANCOL. The values for 2015 in the IPUMS variable now correctly reflect the values in the underlying data.
- To accommodate a change in the shingles vaccine variables available from NHIS introduced in 2018, the previously-available variable SHOTSHINGLE is now SHOTZOSTEV.
- We have replaced the value of MORTWT with PERWEIGHT for the 1986 sample only but set the value of MORTWT to 0 for persons not eligible for mortality follow-up (MORTELIG values not equal to 1). In 1986, the NCHS did not adjust the original NHIS person weight to control for selection bias and instead has advised users to use PERWEIGHT in analyses of the 1986 mortality data. IPUMS NHIS has replaced the value of MORTWT with PERWEIGHT to simplify correct weight selection for users.
July 2, 2019
- Newly released 2018 NHIS data from NCHS are now integrated into the IPUMS NHIS database.
- The universe information for WORFEELEVL was incorrect in 2010-2017. It previously indicated that all sample adults selected for the relevant disability supplements were asked this question; the current universe information reflects that a person's responses to WORFREQ and WORRX affect whether or not the person answers WORFEELEVL. The codes and data for this variable have not been modified; only the universe documentation.
- The universe information for PAINFEELEVL, WALKDIF5BL1, and WALKDIF1BL1 were incorrect in 2011-2017. The previous universe information did NOT specify that these variables were only asked of sample adults receiving the family disability questions or the adult functioning and disability supplement who met other criteria for the variables (e.g., had pain some days, most days, or every day, or refused to answer or didn't know how often they had pain in the past 3 months for PAINFEELEVL). Previous documentation correctly captured topically-specific criteria for these variables; only information about the disability supplements was excluded. The codes and data for this variable have not been modified; only the universe documentation.
- The universe information for WALKDIF1BL2 and WALKDIF12ST1 were incorrect in 2010-2017. They previously indicated that all sample adults selected for the relevant disability supplements were asked each question; the current universe information reflects that the universe was further restricted by persons who use equipment or receive help getting around but do not use a wheelchair or scooter. The codes and data for these variables have not been modified; only the universe documentation.
March 11, 2019
- The linked NHIS-NDI mortality data updated for respondents in the 1986-2014 NHIS surveys is now available through the IPUMS NHIS site. Data from NDI are through December 31, 2015. For more information on the linked NHIS-NDI data, see technical documents from NCHS on the creation and analysis of the NHIS Linked Mortality Files.
- Previously, BALWHEARING erroneously referred to the NHIS variable CBALBHD, which associated headaches with balance problems. BALWHEARING now correctly refers to the original NHIS variable for hearing changes or problems associated with balance problems.
- IPUMS NHIS previously assigned values to NIU for members of the 1992 Hispanic oversample for all mortality variables available for the NHIS-NDI linked data updated through 2009 survey respondents and 2011 deaths. In the release of the linked NHIS-NDI data updated through 2014 survey respondents and 2015 deaths, we have reversed this decision and values of mortality variables for members of the Hispanic oversample retain their original values. Users who wish to know more about the implications of the Hispanic oversample for their analyses should refer to our user note on the 1992 Hispanic Oversample.
- In a previous version of MORTCMS, IPUMS NHIS staff used information from the linked NHIS-NDI file containing information about the vital status of survey participants through December 31, 2011. However, MORTCMS was not indicated as a source of mortality information except for 1994-1998 NHIS survey participants and values of “yes” have not been updated for these participants after December 31, 2006. By using information updated through December 31, 2011, thus increasing the number of “NIU” cases, we misrepresented the denominator for this variable. To correct this mistake, we have revised MORTCMS to use information from the linked NHIS-NDI file containing information about vital status of survey participants through December 31, 2006.
November 27, 2018
- In hierarchical data extracts, the variable NUMPREC incorrectly only listed values of 1 and 2 on person records, instead of matching the value as reported on household records. This has been corrected; household- and person-record values of NUMPREC now match.
- The variable PERNUM is now included on injury-level records to support hierarchical IPUMS extracts in R.
August 16, 2018
- Data from the 2017 Imputed Income files are now available through IPUMS in addition to variables from the following supplements: Complementary and Alternative Medicine, Cultural Competence, Epilepsy, and Heart Disease and Stroke.
- The codes for the variable EDUC were updated. The labels and frequencies associated with each label remain the same, but the codes/values associated with these labels have been modified. The former codes were "00: NIU"; "01: Never attended/kindergarten only"; "02: Grade 1"; "03: Grade 2"; "04: Grade 3"; "05: Grade 4"; "06: Grade 5"; "07: Grade 6"; "08: Grade 7"; "09: Grade 8"; "10: Grade 9"; "11: Grade 10"; "12: Grade 11"; "13: 12th grade, no diploma"; "14: High school graduate"; "15: GED or equivalent"; "16: Some college, no degree"; "17: AA degree-technical/vocational/occupational"; "18: AA degree-academic program"; "19: Bachelor's degree (BA, AB, BS, BBA)"; "20: Master's degree (MA, MS, Med, MBA)"; "21: Professional (MD, DDS, DVM, JD)"; "22: Doctoral degree (PhD, EdD)"; "97: Unknown-refused"; "98: Unknown-not ascertained"; "99: Unknown-don't know".
- The variable IRCAUSE was asked only for injuries in 2004-2016; however, in earlier years, it was asked of injury and poisoning episodes and persons could report "poisoning" as the cause of the injury/poisoning episode. IPUMS has assigned all cases classified as poisoning episodes to have a cause of "poisoning" in the variable IRCAUSE in 2004-2016.
- The variable IRPOISYN incorrectly reported that there were no poisoning episodes in 2014; this has been corrected. Other variables related to poisoning episodes are not affected by this change.
July 9, 2018
- Newly released 2017 NHIS data are now integrated into the IPUMS NHIS database.
- Beginning in 2004, sample children who had surgery can report how many surgeries (both inpatient and outpatient) they have had. This information was appropriately included in the variable SURGERYRNO. However, it was also inappropriately included in SURGRYROUTNO, which reports the number of outpatient surgeries for sample children. The source data is now only offered through SURGERYRNO; additionally, sample adult data was added to SURGERYRNO for 2016.
- The codes and value labels for WHYNOUSLPL. The code "06: Other reason" has been modified to "31" and more clearly identifies that "other" can be interpreted as a reason other than those outlined in codes 1-5 of WHYNOUSLPL. Additionally, codes "98: No usual place, unk why" and "99: Unknown if has and why" have been modified to "91: No usual place, reason missing and "92: Both usual place and reason missing," respectively.
- The codes and value labels for TYPPLSICK have been updated. The code "56: Other places" has been modified to "57". This only affects 1997-forward samples.
April 4, 2018
- The coding schemes for FBNHCAN, FBNUM, FSNHCAN, and FSNUM have changed to reflect the original NHIS coding scheme whenever possible. Please refer to the "codes" tab on each variable's page for more information.
- In 1982-1989, the variable HINOUNEMPR classified persons who indicates some type of general insurance coverage as having an "Unknown" response. These cases have been reclassified as out of universe.
- In 2004 and the first two quarters of 2005, only persons who did not answer "Single Service Plan" when asked about insurance coverage (HISINGLE) were asked a probe question about having additional single service coverage (SSPROB). However, beginning in quarter 3 of 2005, persons who reported single service plans in the initially health insurance coverage question were automatically assigned as responding "yes" to SSPROB. To increase consistency over time, IPUMS NHIS has recategorized persons who report having single service coverage (HISINGLE) in the initial health insurance question, as responding "yes" to SSPROB in 2004 and first two quarters of 2005; previously these cases were coded as NIU. Additionally, in all years, IPUMS NHIS has recategorized the limited number of "unknown-not ascertained" responses to HISINGLE as also being "unknown-not ascertained" in SSPROB, rather than categorizing them as NIU.
- The codes for INCFAM07ON were changed and expanded to be two digits wide; this new scheme allows for income groups with narrower, defined ranges to share a leading digit with income groups that are broader or lack one endpoint. The underlying variables, values, and value labels have not changed.
- We recently learned that in October of 2014, NCHS re-released the person-level files for 2012 and 2013 because of errors in two health insurance variables (HIP1MDREQ and HIP2MDREQ). IPUMS NHIS has re-released these two variables for 2012 and 2013 using the corrected NCHS files.
- The previous codes and labels for DISTESTFLG incorrectly specified that all sample adults selected to receive the family disability questions (FDB) in 2012 also completed the adult functioning and disability (AFD) questions. Approximately one half of sample adults who were selected to receive the FDB were also selected to receive the AFD questions. The previous version of this variable was not incorrect for identifying recipients of FDB questions in 2012, but was mislabeled and did not accurately identify whether or not sample adults who received FDB questions also received AFD questions.The codes for DISTESTFLG have been modified and now correctly differentiate between sample adults who received both FDB and AFD, and those who received FDB only. The frequencies have only changed for 2012, though the codes have been modified for all years of data. Additionally, in 2010, DISTESTFLG labeled families who received FDB questions as receiving ADB/CDB questions; no ADB/CDB questions were asked in 2010. These cases have been correctly recoded as having received FDB questions.
- In 2010, the previous codes for FAMRESPFLAG incorrectly assigned both "not family respondent" and "family respondent" values to "family respondent." IPUMS NHIS staff have corrected this error and re-released FAMRESPFLAG for 2010.
November 8, 2017
October 19, 2017
- Data from the 2016 Imputed Income file and the Balance, Vision, Tobacco and E-cigarette Use, and Diabetes supplements have been added.
- IPUMS NHIS corrected source data for the integrated variable BALCHRYST. BALCHRYST was incorrectly using source information from the variable BALCVERT in the 2012 sample.
- IPUMS NHIS updated the integrated variable BALSYMPTNO to correct an error. Previously, NIU cases were erroneously represented with a code of 7 in the 2008 sample.
- The codes for "some day" and "every day" were flipped for the variable ECIGED. In the re-released data, a code of "1" refers to "every day" use while a code of 2 refers to "some day" use; these labels were attached to the same codes in previous releases, but the code of "1" was actually reporting "some day" use and "2" was reporting "every day" use. This affects 2014-2016 samples.
- In reviewing the modified 2016 flu vaccine variables for sample adults, IPUMS NHIS staff discovered that the sample child variables had undergone a similar modification in 2011 without thorough documentation by IPUMS NHIS. While older versions of flu vaccine variables asked about flu shot vaccination and flu spray vaccinations separately, later years ask about flu vaccinations generally with a follow up question about the mode of vaccination. As such, all flu vaccine variables have been revised. This includes renaming variables for consistency and clearer indication of what mode of flu vaccination they report, as well as using special programming to extend availability of variables that are no longer offered on the public use NCHS files, or replicating the newer versions of these variables in older sample years. A list of deleted variables and their replacements can be found here.
- A fix corrected the programming used to create POVIMPHHS2 in 2014. Previously, the programming compared the family's imputed income against the federal poverty line for a family of one, regardless of family size. The fix has updated the programming so that family size is correctly accounted for in assigning poverty status.
- IPUMS NHIS changed the integrated variable name of RXBALPROB to BALRXPROB. This change was made to improve the searchability of this variable within the Balance group.
- To enhance comparability between IPUMS NHIS and other IPUMS data projects, IPUMS NHIS staff changed the names of several integrated income variables. These include GOTINTAMT (now INCINT), GOTINTAMTFL (now QINCINT), GOTDIVAMT (now INCDIVID), GOTDIVAMTFL (now QINCDIVI), GOTOTHPNAT (now INCRETIR), GOTOTHPNATFL (now QINCRETI), GOTWELFAMT (now INCWELFR), GOTWELATFL (now QINCWELF), GOTOTHERAMT (now INCOTH), and GOTOTHATFL (now QINCOTH).
July 26, 2017
- Newly released 2016 NHIS data are now integrated into the IPUMS NHIS (formerly IHIS) database. This data release also includes a handful of variables that were either new in 2015 or existing variables that have now had the 2015 sample added.
- The variables HOMECARE2W and PHONEMED referenced the wrong underlying variable and only listed "no" values in the 1997-2015 samples; the underlying variable for the 1997-forward samples has been updated. Pre-1997 values are not affected.
- The codes for the variable RELATE in 2015 were coded to the incorrect relationships other than "Householder", "Spouse", and "Unmarried partner"; 2015 codes have been updated and now report the correct relationship to the household reference person.
- IPUMS NHIS has changed the procedure in which the variables INDSLH1, INDSLH2, OCCUPLH1 and OCCUPLH2 (*LH variables) are created. In 2010, these four variables had complementary, but mutually exclusive universes with INDSPY1, INDSPY2, OCCUPPY1 and OCCUPPY2 (*PY variables). The 2010 values from the *SPY variables were combined with the *SLH variables in order to increase comparability with the 2015 values. Please see the variable descriptions for additional detail.
- A fix corrected the underlying variable for CIGMENTHOL in 2015 and correctly reassigned the labels; previous versions of the data inverted the labels for "Plain" and "Menthol" in 2005, 2010, and 2015.
- A fix restored YRSINUS for 1998. In previous versions of the data, all values of YRSINUS for 1998 were missing.
- A fix restored SUPP3WT for 1994 and 1995. In previous versions of the data, a programming error set all values of SUPP3WT for 1994 and 1995 to zero.
May 31, 2017
- Data from the 2015 Cancer Supplement are now available. Users should note changes in the 2015 family cancer history variables as compared to earlier Cancer Modules, especially in the omission of many specific types of cancer and the combination of colon and rectal cancer into colorectal cancer and cancers of the head, neck, and mouth into a single head and neck cancer set of variables.
September 7, 2016
- Imputed income variables from the 2015 NHIS data are now integrated into the IHIS database.
- Newly released 2015 NHIS data are now integrated into the IHIS database.
- More than 1,000 newly integrated variables from NHIS historical supplements on Disability (1977, 1994, and 1995), Polio (1994 and 1995), Youth Risk Behaviors (1992), Diabetes (1989), Hypertension (1974), HMOs (1975), AIDS (1987-1995), and Healthcare Access and Utilization (1993-1996, 2001) are now available.
- MORTWT and MORTWTSA were updated to remove implied decimals. This does not affect underlying data, but would affect control files and formatted data extracts.
- MORTWTSA was widened from 7 to 8 columns, but because the maximum value of the underlying data never exceeds 7 columns, there is no change in the value of MORTWTSA.
- HYPLEVEL was rearranged to an ordinal arrangement of low/normal/borderline/high blood pressures. In addition, the previous version incorrectly assigned the same output code to normal and high blood pressure. These are now correctly assigned separate codes.
- In previous versions of HINOTCOVE, a value of 1 indicated "Not covered" and a value of 2 indicated "Covered" HINOTCOVE has been updated for internal consistency, with a value of 2 indicating "Yes, has no coverage," and a value of 1 indicating "No, has coverage." Proportions of uninsured persons have not changed.
- In previous versions of HINOTCOV, persons in 1980 who had unknown values for GOTSSINOW or GOTAFDCPROG were automatically assigned unknown values for HINOTCOV. This programming error resulted in 239 persons incorrectly assigned unknown values for HINOTCOV in 1980 only. An additional programming error resulted in 409 person in 1976 being incorrectly identified as not insured. They are now correctly identified as insured.
- In the most recent version of HIPUBCOVE, a programming error assigned values of "Unknown/don't know" to all persons without public coverage and to certain persons with coverage under HIOTHGOVE, HISTATEE, and HICHIPE. This error has been corrected. Data users who created extracts with HIPUBCOVE prior to 9/18/2016 should re-submit extracts for correct data.
- In previous versions of HIPUBCOV, 6,369 persons in 1992 who had unknown values for other public insurance coverage were incorrectly assigned "No" responses. Those persons have been reassigned. In addition, 590 persons in 1980 incorrectly assigned values of NIU have been reassigned values of "Unknown/refused."
- Incomplete universe information was updated for WORRX, ARMFTIM5, LAMEMORCON, WALKDIF12ST1, WALKDIF1BL2, LAWALKCLIMOTH, DISWALKP, TIREDURALAST, VAC2FLUYRMO, FLRETNO, DEPFEELEVL, MCARPLUS, GOTNONSSDIS, TYPPLSICK, HINOTCOV, HRAUSESQOL, WALKDIF5BL2, IRIPSTU, HICHAMPANY, WALKDIF1BL1, FLSURGRY, VACFLUYRMO, HIPCONAFFORDR, VDIFAMOUNT, INJDONA, FLNEC2, HIMILVA, WALKDIF5BL1, HIPTYPER, INJDOREF, LARA2LITRDIF, CLIMANEC1NO, LAMEMDIFAMT, LAWALKCLIMDIF, HRPROBAMT, GLASSES, and IRIPEMP.
- Added variable descriptions for INDSTRN204, WHYNOWK2, STATEPREM, MCAIDPREM, OTHGOVPINC, MCAIDXCHG, OCCUPN104, OTHGOVPREM, MCAIDPINC, CHIPXCHG, CHIPPINC, OCCUPN204, EXCHANGE, OTHGOVXCHG, INDSTRN104, and CHIPPREM.
- For 2014, users were incorrectly advised to use SAMPWEIGHT, rather than PERWEIGHT for the person-level injury variables (e.g., INJPLANA). The weighting information for all person-level injury variables has been corrected.
- IHIS has enabled the Attach Characteristics feature in the extract system. Attach Characteristics allows users to write the values of selected characteristics of co-resident mother, father, and/or spouse to own record. This feature is also available for same-sex partners. Attach characteristics uses information in the MOMLOC, MOMLOC2, POPLOC, POPLOC2, and SPLOC.
March 30, 2016
- Nearly 1,000 variables covering the Alcohol Use, Occupational Health, Mental Health, Family Resources, Adult Supplemental Items, and Health Insurance Exchange supplements are now available from the IHIS online database.
- This release also incorporates spouse person number SPLOC and method for assigning spouse person number SPRULE, the first of a set of expected Family Pointer variables that define family relationships between household members in a comparable way across the 1963-present time period.
- IHIS updated the variables IRBODY1 through IRBODY4 in 2007 to correct an error. Previously, nose injuries were being incorrectly recoded as "head (not face)" injuries.
- IHIS corrected source data for the variable VISITYRNO for 2011-2014. VISITYRNO was incorrectly using source information for the number of home visits (rather than the number of office visits) for sample adults, although it was using the correct information for sample children and for other sample years. It has been updated to refer to the number of office visits in 2011-2014 for both sample adults and sample children.
- IHIS corrected the formula used to calculate the variables POVIMPHHS1 through POVIMPHHS5 for the 2011-2014 samples. IHIS staff create POVIMPHHS1-POVIMPHHS5 using FAMSIZE and INCIMPPOINT1 through INCIMPPOINT5 to generate poverty ratios using the imputed income data that mimic the poverty ratios created by NCHS using unimputed income data. In the prior release, IHIS staff mistakenly used the base and increment values for calculating poverty thresholds from the current rather than the previous year. Base and increment values are set by the U.S. Department of Health and Human Services.
January 15, 2016
- The integrated variable CLIMRETMO previously referred to the incorrect original NHIS variable (LUNIT14A) for the 2011-2014 samples. It has been updated to refer to the correct original NHIS variable (LDURB14A).
- The integrated variable HI1PREINC previously referred to the incorrect original NHIS variable (PLNPAY71) for the 2013-2014 samples. It has been revised and now refers to the correct NHIS variable (PLNPRE1).
- BMICALC previously only had values for NIU and missing cases for the 2015 sample. It has been corrected to include the full complement of calculated BMI values for the 2015 sample.
- Incorrect variable names for variables with the original NHIS codes were corrected in the description for the variable ELDCH.
- Incorrect references to a non-existent variable were replaced with information about an existing variable in the description for the variable OCCUPN104.
- Incorrect references to a non-existent variable were replaced with information about an existing variable in the description for the variable INDSTRN104.
September 23, 2015
- Imputed income variables from the 2014 NHIS are now harmonized into IHIS. 2014 survey text is now available, allowing users to easily access the original survey questions associated with integrated IHIS variables.This release also includes over 500 variables from historic NHIS supplements including immunizations, arthritis, assistive devices, home care, and more.
- In September of 2015 NCHS re-released the 2014 injury and poisoning episode file. This re-released file included variables created through medical coding, including the ICD-9 codes for conditions associated with each injury/poisoning record, and external cause codes (or E-codes). IHIS now includes these variables as well as the associated person-level variables. There were additional changes in frequency distribution between the original and re-released injury/poisoning episode-level files; these differences also affect the person-level variables created from the episode-level data. These changes are not systematically associated with a single variable or group of variables. Users who created an extract using 2014 person-level or episode-level injury data prior to September 23, 2015 should be advised that the data have changed and the current IHIS version reflects the most recent NCHS release of the injury/poisoning episode file released on September 9, 2015.
- This release also incorporates the re-released NCHS files with corrected child birth weights for the 2004-2010 sample child files. The originally released 2004-2010 sample child birth weight values contained few high birth weight children and an elevated number of very low birth weight children. After re-examining the raw data, NHIS staff identified an error in processing the raw two-digit data for pounds, where values were accidentally collapsed to a single digit and obscured high birth weight children; these errors were transferred when creating the parallel variable that reports birth weight in grams. Additionally, the top and bottom codes for persons reporting child birth weight in grams misclassified children with birthweights over 5485 grams into a missing category. Ounce data were not originally incorrect and are therefore not affected. IHIS integrated these corrected birth weight files and updated the affected variables BWGTLB and BWGTGRAM, as well as offering a new summary variable BWGTTOTOZ, which reports birth weight in total ounces. Users interested in comparing the original birthweight files should contact IHIS user support to obtain the previous, incorrect values for these years. Original NCHS documentation on the corrected birth weight files is available here.
- IHIS updated the variable MARST in 2014 to disaggregate the persons who reported being "Married" into "Married-spouse present" and "Married-spouse absent". This change only offers additional detail and it does not change the marital status of persons.
- Four variables providing information on surgical insurance coverage (SURGBLUE, SURGPREP, SURGPRIV, and SURGUNKN) were updated to disaggregate persons not in universe from no responses; this involved a change in codes associated with each response category. The variable SURGBLUE74 was eliminated and combined with SURGBLUE.
- IHIS corrected source data for the variable HIPUBCOVE for all years. NCHS creates an alternate version of insurance coverage variables which are back-edited based on verbatim responses to health insurance plans named by respondents. Previously, HIPUBCOVE mistakenly duplicated the directly reported information about different types of public coverage available in HIPUBCOV. HIPUBCOVE now correctly summarizes information from the NCHS-edited versions of variables that report if an individual is covered by any of several public programs. Detailed coverage of the NCHS back editing process for health insurance variables can be found in the IHIS Summary and Key Insurance Variables user note.
July 20, 2015
- IHIS now offers most variables from the 2014 NHIS, including measures of health care access, health behaviors, disability and conditions, core demographics, injury and poisoning episodes, and socioeconomic indicators.
- IHIS changed the codes and updated the value labels for the variable LVISOTHTYPE to account for new response categories included in 2014. The underlying frequencies remain the same for all previously available years; however the numerical codes associated with response categories have changed slightly.
- IHIS updated the missing data codes the variable INCFAM07ON to account for changes in the source data in 2014. The underlying frequencies remain the same for all previously available years; however the numerical codes associated with missind information response categories have changed slightly.
June 29, 2015
- IHIS now includes the updated mortality files, with data from 1986-2009.Variables from these files allow users to link 1986-2009 IHIS variables with death certificate information from the National Death Index through December 2011, enabling statistical analyses of the relationship between mortality and the rich array of socio-demographic characteristics, health factors, and health care access information available in the NHIS data.
- IHIS has created harmonized industry and occupation variables for 1969 to the present to allow for comparisons of consistent categories time. The original detailed industry and occupation codes are also now available in the same variables over time (IND and OCC), making it easier for users to include all relevant information in their extracts. For more information, please see the
- This release also integrates the injury and poisoning episode records for 1997-1999. These variables were originally released by the NHIS as two separate files, but IHIS has combined the two files into a single record type to improve ease of use and comparability. See our Injury and Poisoning Supplement User Note for more details on this harmonization or use the variable IRPOISYN to identify records that were originally on the poisoning file in 1997-1999.
- Additionally the release adds historical supplements on the topics of hearing, vision, asthma, unintentional injuries, as well as pregnancy and smoking.
- IHIS is also debuting its data brief series, a set of concise research reports that highlight unique aspects of the IHIS database and interesting research topics that can be investigated using the data. Brief 01: Multigenerational Families and Food Insecurity in the United States, 1998-2013.
- IHIS staff updated the codes for the variable
- The variable IRHOSP was eliminated after being merged with the variable IRMEDHOSP as they contained parallel information available in different years. The frequencies, codes, and meaning of the variable remain the same.
- The variable NHISIID was updated to be 19 columns wide. This was necessary so the unique identifier could include IRPOISYN in 1997-1999, which distinguishes injury episodes from poisoning episodes in the original NHIS files. IRPOISYN is also necessary to identify if the value in IRINJNUM is the nth injury record or the nth poisoning record of the person in the 1997-1999 samples.
- The variable POISCENHEAR was corrected by removing children under age 10 from the universe into a separate variable for children, POISCENHEARCH. Children under age 10 were previously included with sample adults, leading IHIS software to inappropriately assign the weight PERWEIGHT to sample adults. POISCENHEAR now has a correct weight of SAMPWEIGHT and POISCENHEARCH has a correct weight of PERWEIGHT.
April 28, 2015
- The IHIS now includes imputed income data for 1990-1996. While these differ in structure from the 1997-forward imputed income files, they can be used to fill in missing income data for all persons in the 1990-1996 samples. Over 2000 variables from the complementary and alternative medicine (CAM) supplements for sample adults and sample children have been fully integrated into the IHIS data for the 2012 sample. A change in the structure of the CAM questionnaire makes it difficult to compare original NHIS CAM data from 2012 with previous samples; the newly released IHIS data include the original variables released by NHIS in 2012 as well as variables constructed through special programming that are comparable to CAM variables released by the NHIS in previous years to allow for easier analysis of these variables over time. The release also includes a series of variables rom the 2002 survey on barriers to community participation. Finally, historic child development and child medication variables from the 1981 child health supplement are now available through the IHIS.
- IHIS briefly offered updated NHIS-NDI linked mortality files for the years 1986-2009. These data were from the National Center for Health Statistics files released in February 2015. However, on April 28, 2015 NCHS re-released the files citing a problem with the public use files originally offered. The IHIS data originally released utilized the since removed NCHS public use files that contain errors; therefore, IHIS has restored its previous version of these variables, which provide mortality follow up data for 1986-2004. IHIS will prioritize processing and harmonizing the corrected mortality-linked files for 1986-2009 for inclusion with its next data release.
- IHIS staff found and corrected a typographical error in the input data for the variable FAMDELAYCONO in 2009.
- The codes and labels for RELMOM were updated to reflect the inclusion of a new response category in 2013. Prior to 2013, biological and adoptive relationships to the mother were categorized separately. However, beginning in 2013 these were condensed into a single category. All codes with values of greater than 10, or anything other than "NIU" or "Biological", are off by one category (i.e., "adoptive" is mistakenly coded as "biological or adoptive", "step" is mistakenly coded as "adoptive", "foster" is mistakenly coded as "step", etc.) in any extracts made after June 28, 2014. The error has been corrected. Values for RELPOP and RELSIB were and remain correctly coded for this change.
- The frequencies for SINGLE were updated to more accurately reflect the universe of all persons in 1989. Cases previously classified as "NIU" have been recoded as missing values.
- The variable SDENTAL was incorrectly coding all persons to "NIU" in 1993-1996. This has been corrected and the variable now reflects the correct codesa nd frequencies for these years. No other years were affected.
- The language in the variable description for HICHANGEYR included misleading language. The description has been updated to reflect that a "yes" response indicates that a person retained the same type of insurance coverage for the past 12 months.
- The value labels for the 2007 complementary and alternative medicine (CAM) variables that report the most important condition a person treated using CAM therapies were updated. The labels and frequencies have not changed; but the numeric output codes have been updated for those therapies also available in 2012 to improve comparability over time. Affected variables are ACUTCONMOST, BIOTCONMOST, COMTCONMOST, DITTCONMOST, EHTTCONMOST, FOKTCONMOST, HER1TCONMOST, HER2TCONMOST, HOMTCONMOST, HYPTCONMOST, MASTCONMOST, MOVTCONMOST, NATTCONMOST, RELTCONMOST, and YTQTCONMOST.
March 2, 2015
- The 2013 supplements not included in the October 2014 IHIS data release have been integrated. These include 2013 variables on asthma, cancer screening, health care access and utilization, disability, epilepsy, food security, heart disease and stroke prevention, hepatitis, immunization, immunosuppression, internet access and email usage, mental health, mental health services, and tobacco. IHIS is also releasing a series of historic variables with information about HIV/AIDS testing; these include variables on reasons for being tested, reasons for not being tested, locations of testing, information about testing, and discussion of test results. Additionally, this release provides full documentation for an additional 1,250 variables on topics including exercise, weight control, chronic mental illness, colon cancer testing, insurance, public health service tests, quitting smoking, mammograms, Pap testing, hypertension, income, and HIV/AIDS information and testing. All integrated IHIS variables now have full documentation, including a description of the variable, an overview of the codes, universe information, and a discussion about comparability of the variable over time. Survey text is now available for integrated 2012 and 2013 IHIS variables. Additionally, the original NHIS person-level linking keys (HHX, FMX, and PX) for the 1997-forward samples and day of interview (INTERVWDAY) for the 1970-1996 samples are now available.
- To better serve users, IHIS upgraded the system that processes data made available on its website. As part of this process, IHIS did a thorough inventory of existing variables and programming, uncovering typographical errors as well as opportunities for improving comparability. Because of the more thorough review of our variables, there are more revisions than usual. These changes are documented below and arranged by variable topic.
- Health Insurance Variables
- Certain summary and key health insurance variables were modified for cases of partially missing information. For variables that summarize information across multiple plans with partially missing information, IHIS now prioritizes the non-missing information when coding these cases. Where persons who reported not having a certain coverage feature for a plan, but also reported being uncertain about their coverage by that plan (or have missing data for coverage by the plan) were previously coded as unknown/missing, these persons are now coded as "no" responses, giving priority to their known "no" response to the specific plan feature over missing information on a less informative variable. For example, the variable HIPMDOPR reports if a person has any private insurance plan that pays any costs for a doctor not on the plan's preferred list. HIPMDOPR has been reconstructed so persons who definitively reported that their private insurance plan would pay for this but have missing data on if they are covered by this plan are now coded as responding "yes" instead of missing. Variables affected by this include HIPBUYOWNR, HIPDENTCOVR, HIPMDLISR, HIPMDOPR, HIPMDPICR, HIPTYPER, HINOUNEMPR, and HIPWORKR in all available samples. Additionally, HIPBUYWONR was updated to address "unknown" cases that were mistakenly being coded to "no".
- Where possible, more nuanced missing codes were made available for variables. Users may notice that where some health insurance variables previously had missing codes of 9 only, there are now codes specifying "unknown-refused" (7), "unknown-not ascertained" (8), or "unknown-don't know" (9)
- Additionally, typographical errors were corrected in the programming for the following variables, resulting in changes to the output codes: HIPDENTCOVR for 2004-2013; SINGLE in 1993-1996;
- The missing codes for the variable HIP4MDPIC in 1992-1996 were updated to create coding consistent with other variables. Only the codes were changed; the underlying frequencies and data remain the same.
- The previous programming for HINOTCOV in 1992 previously eliminated all legitimate observations; correct programming has been restored.
- The previous programming for HINOTCOV in years other than 1992 prioritized incomplete data cases over known cases; the new version of this variable gives precedence to persons who specifically report a known lack of coverage (output code of 2 or "yes, has NO coverage"), persons who have known coverage (output code of 1 or "no, has coverage"), followed by unknown/missing observations. Because this variable asserts that persons have NO coverage, cases with missing data are coded conservatively and continue to receive unknown codes.
- The previous programming for HINOTCOVE in 1997 incorrectly assigned all values to either 2 "Covered" or 8 "Unknown - not ascertained." The programming has been updated to produce the correct assignment of values.
- The variable HIPTYPER was updated to consistently prioritize having any HMO plan; the variable description reflects this change.
- Pre-1997 conditions variables from condition records
- The coding scheme for pre-1997 person-level conditions variables, created from condition-level records, was modified slightly. The previous scheme assigned information from only the first condition record with each condition code (e.g., ICD-9 code), regardless of how the condition record was generated. The modified coding scheme searches through all condition records for each person and prioritizes any information provided in direct response to a conditions survey question. This affects a very small number of cases, changing responses from "no reported conditions" (codes of 1 or 10) or "condition indicated somewhere other than a direct response to a conditions survey question" (codes 22, 212, 222, 232) to "Yes, indicated by a response to a direct conditions survey question."
- The variable FEMTROUBYRC was also modified to prioritize more specific sub-conditions included in this variable. The variable now prioritizes any infertility records in the coding scheme, followed by inflammatory disorder of the female pelvic organs, then other disorders of the female reproductive tract (previous coding scheme first categorized inflammatory disorders, then other disorders, then infertility). This affects only a marginal number of cases where persons reported multiple condition records related to problems with female reproductive organs.
- IHIS staff found an error in the linking keys for the 1983 and 1992 conditions records. In 1983, no persons were previously reporting "yes" responses to any conditions because of this error; the 1983 conditions frequencies have since been corrected and correctly report the conditions experienced by persons. In 1992, repairing these linking keys allowed for the inclusion of additional conditions records. No records were eliminated or changed from "yes" to "no" in this correction; only additional records have been included.
- Complementary and Alternative Medicine (CAM) variables
- IHIS staff found and corrected an error in the source data for the variable YTQEXNO. Frequencies have been corrected.
- The variable VITPAIDL, created using special programming, was modified to better specify a single marginal case where, in the original NHIS data, the person reported an unknown value for frequency and amount spent purchasing vitamins, but was classified as "not in universe" in the corresponding frequency time unit NHIS variable.
- The variable MASPAIDT was updated to better specify marginal cases where persons report missing information about insurance coverage for any or all of the part of their massage treatment costs, but provide a legitimate response to the variable asking for the out-of-pocket amount spent. These cases were previously classified as "not in universe", but have been modified to report the total amount the person reported spending out of pocket on massage therapy.
- Sample children were inadvertently excluded from the variables EHTEV, FOKTCANC, DITPRITEV, MOVFELD, and MOVPILAT in 2012. The sample adult frequencies remain the same; sample children are now included.
- Other variables
- IHIS staff found an error in the codes for the variable TPASTEWHITE. The categorical labels for these codes have been updated; previous labels erroneously referred to meeting ADA and FDA standards, but the underlying data only refer to marketing strategies of toothpastes.
- The IHIS variable STRATA previously classified a limited number of "not in universe" observations inconsistently in 1969-1984. These cases have been correctly given values of 0 to indicate their "not in universe" status; these STRATA cases were previously recoded to 1000 in 1969-1972 and 2000 in 1973-1984.
- Because of different rounding conventions in programming languages, a limited number of cases for the variable HHWEIGHT in the 1974-1996 samples and BMICALC in all samples were previously rounded down in instances where the computation of HHWEIGHT ended in exactly .50; while a byproduct of how different programming languages round, it is mathematically incorrect. The new data conversion program correctly rounds these cases up.
- A typographical error in the input data for the variable IMPFAMTCFLAG was corrected in 2009 and 2010.
- A typographical error in the classification of not in universe codes was fixed for the variable DIFSCORE in 2001, 2003, and 2004. Values previously coded as 97 are now correctly displayed under the expected NIU code 96.
- A typographical error in the programming for the variable SMOKEALARM was corrected in 1994; "no" and "yes" responses were inadvertently shifted up one response category, where "no" responses were mistakenly listed as "yes", and "yes" responses were mistakenly listed as "unknown-refused". Frequencies and codes have been corrected; this only affects 1994.
- Nineteen "not in universe" cases for the variable OPOUTPAT2 in 1981 were mistakenly being classified as 96 rather than 0. While 96 sometimes represents a "not in universe" code in other IHIS variables, for OPOUTPAT2 the "not in universe" code is 0 and 96 represents "replacement and removal of therapeutic devices."
- A typographical error in the source data for the variables IRECODE1, IRECODE2, and IRECODE3 were found and fixed in the 2013 sample. These variables are now corrected; these typographical errors affected only the injury-level variables; the person-level injury variables are NOT affected by this change.
- The injury-level variable IRDAYSLB in the 2013 sample was incorrectly specifying all values as 0 because of a typographical error in the special programming; this has been corrected and lower bound values for number of days since the injury/poisoning episode are now being correctly reported.
- A scientific notation character from the original data in the variable IRDAYSUB for the 2013 sample was handled incorrectly in previous special programming. Three cases where output codes previously read 1003 or 2003 have been corrected to 1000 and 2000, respectively.
- An error causing incorrect value labels for two variables that report the year of an event were updated. In the 2010 sample, the variables HYSTDYR (year of hysterectomy) and SKNXMDYR (year of most recent skin cancer exam) incorrectly labeled the repsonse categories "2007", "2008", "2009", and "2010" as "not in universe". The labels for these values now report the correct year.
- The codes for the variable PAPLTOLDYN were incomplete in the 2010 sample. "Only told if there was a problem" (3) was added in; these values had previously been assigned to "Unknown-refused" (7). Missing codes were correspondingly misassigned as well; all missing codes have since been updated to correctly reflect the reason for the missing data (i.e., "refused", "not ascertained", "don't know").
- A number of variables contained incomplete information for the translation of original data codes; a systematic review was conducted and all instances have been corrected. The following variables are affected by the correction: ALCDAYS2WK, BSNUM, CNCERVAG, COKUSWAL, CSOFFGR, CXRAYREX, EHDOTHER
October 6, 2014
- This release includes access to the new IHIS hierarchical extract system, which allows users to select episode-level data for injuries and poisonings as well as person- and household-level data. For more information on these episode-level data, please see the Injury and Poisoning Supplement User Note.
- IHIS released 2013 imputed income variables. New 2013 variables and variables from reoccurring supplements not available in 2012 will be released sometime in fall 2014.
- IHIS added full documentation for more than 1250 variables currently available on the IHIS website on topics including: child health conditions, digestive conditions treatment, foot problems, housing modifications, insurance, limitation, women's health, and work exposure.
- In response to a user query, IHIS has corrected an error in the response categories for RELMOM. When updating response categories to reflect the concatenation of multiple response categories in 2013, response categories for previous samples were not all updated accordingly. This will affect frequencies for 1998-2012 for the categories of stepmother, foster mother, mother-in-law, unknown-refused, and unknown-not ascertained. This error only affects data extracts made between July 31, 2014 and September 17, 2014. All extracts prior to July 31, 2014 and frequencies for the "biological" and "adoptive" response categories in 1998-2012 are not affected by this correction.
- IHIS staff found and corrected a typographical error in the data source for the HI2TYPR1 and HI2TYPR2 in the 1993-1955 samples.
- In response to an inquiry from our colleagues at SHADAC, IHIS corrected reversed response categories for the variable SMOKEV in 2013. Persons who made data extracts prior to September 17, 2014 should switch the yes and no codes for the correct values.
- In response to a user query, IHIS re-adjusted the length of the variable NHISPID for 1997-forward. The previous version eliminated a leading zero, creating a variable that was only 15 columns wide instead of the anticipated 16 columns. Users may revise their extracts using the IPUMS extract system to update the value of NHISPID, or modify their existing extracts on their own by inserting a zero into the string value of NHISPID after the first 6 values (i.e., at the end of "00" followed by the year of the survey). While exploring this issue in NHISPID, IHIS staff found and corrected an error in the source data for NHISHID in the 1968 sample. Only the 1968 sample has been modified; all samples reflect the components specified in the IHIS user note on linking.
- After reviewing the Safety Gear and Behavior variables, IHIS made changes to streamline the variables related to seatbelt use. SBELTCHFREQ2 was combined with SBELTCHFREQ to create a single variable that provides data on use of seat belts among children across available samples. The variable SBELT was modified to create a single variable providing data on the use of seat belts among adults across available samples.
- The following variables available in the 1977 sample were updated to create more consistent coding of not in universe values: AMTBEER, AMTLIQUOR, AMTWINE, DRANKFIVE, HASBEER, HASLIQUOR, HASWINE, PROBAVAIL, PROBCOST, PROBGETCARE, PROBHRS, PROBPREVCARE, PROBTRAN, PROBWHERE, HEIGHT, and WEIGHT. In 1977, these variables were asked for a sub-sample of randomly selected persons ages 20 and older who can be identified by the flag SUBSRESP77. Persons selected to respond to these questions who did not complete the questionnaire are placed into different response categories for each question in the original NHIS data. For more detail on how selected sub-sample persons who did not respond are handled in each variable, please see the variable description. Persons wishing to recategorize these non-responding selected subrespondents should use SUBSRESP77 to identify and reassign these persons. The variables asked of this one-third subsample previously indicated that persons should use the IHIS variable PERWEIGHT; after a thorough review of the limited NHIS documentation on these variables, IHIS has released a different weight for these variables: SUB77WT. The Weight tabs for all relevant variable pages have been updated to reflect this change. Other variables affected by this change are BREAKFAST, SNACKS, PHYSACTIVE, WEIGHTK, HRSLEEP, SMOKESTATUS1, SMOKENOW, CIGSDAY.
- IHIS modified the names of several key Insurance variables to improve clarity: HIPRIVATEE, HIMCAIDE, HIMCAREE, HISTATEE, HICHIPE, HIOTHGOVE, HIPUBCOVE, HIMILITE, HINOTCOVE, SINGLEE, SDENTALE, HIHSE, HINDIAN, HIPRIVATE, HIPUBCOV, HINOTCOV, SINGLE, SDENTAL
- HIS has also combined several health insurance variables to reduce redundancy: HI1HMOCOVR, HI2HMOCOVR, HI3HMOCOVR, HI4HMOCOVR, HI1WHO, HI2WHO, HI3WHO, HI4WHO, HI5WHO, HOSPTYPE, HI6WORK, HI5EMP
- For additional information on health insurance variables, please see the IHIS user note on using IHIS Insurance variables
Removed and modified variables.
July 29, 2014
- IHIS released 2013 variables that were also included in the 2012 sample and a newly available variable on sexual orientation (SEXORIEN
). New 2013 variables and variables from reoccurring supplements not available in 2012 will be released sometime in fall 2014.
- IHIS added full documentation for more than 950 variables currently available on the IHIS website on topics including: 2012 Voice, Speech, Language, and Swallowing supplement, street drugs, medical care, select insurance variables, emotional help, child medication, cancer knowledge, and general health.
- A set of variables introduced in 2011 that capture reasons for most recent ER visit, ERLAMBUL, ERLCLOSEST, ERLDRCLOSED, ERLDRSENT, ERLHOSPHELP, ERLNOTHER, ERLSERIOUS, and ERLUSUALPL, were combined with information contained in alternate versions of these variables covering quarter 2 of 2012 through 2013 (ERLAMBULR, ERLCLOSESTR, ERLDRCLOSEDR, ERLDRSENTR, ERLHOSPHELPR, ERLNOTHERR, ERLSERIOUSR, and ERLUSUALPLR). The alternate versions were removed from the IHIS database.
- To accommodate new response categories for family relationship variables RELMOM, RELPOP, and RELSIB, IHIS updated the value labels for these variables. These changes do not affect frequencies.
- In response to a user query, IHIS staff corrected a typographical error in the data source for the variable SAWMENT
in the 2012 sample.
- A typographical error in the data source for the variable HINOFAMR
was corrected in the 2012 sample.
- The not in universe category for variable MOMPNUM
was corrected. The variable reports the person number of the mother in the household. The previous universe statement read "all persons" and categorized "no mother in the household" as a separate response category. The universe now correctly reads "persons whose mother is present in the household". This affects value labels, but not frequencies for the 1998-2012 samples.
- The IHIS online system incorrectly associated SUPP1WT with variables from the family disability supplement in the 2012 sample, when they should have been associated with SUPP2WT. Variables from the family disability supplement are now correctly associated with SUPP2WT and variables from the functioning and disability supplement are now correctly associated with SUPP1WT.
- NHISHID and NHISPID for the 1963-1967 samples were missing the value of year in the first four columns. The latest release corrects this error, so that the values of NHISHID and NHISPID for pre-1968 samples are consistent with 1968-forward samples.
- The universe and codes for DISTESTFLG
were updated for 2011-2012. The previous universe was persons in households that received the family style disability questions, coding persons who received the person-style Adult Disability or Child Disability questions as NIU. Although this was not incorrect, the DISTESTFLG universe has been updated to include all persons and includes flags for both persons in families randomly selected to receive the Family Disability (as well as Adult Functioning and Disability) Questions and persons randomly selected to receive the Adult Disability and Child Disability questions.
Removed and modified existing variables..
June 30, 2014
- The 1997-1999 NHIS included person-level injury and poisoning records, where injury and poisoning episodes were attached to the person level file. IHIS has created person-level variables comparable to those available in 1997-1999 based on the episode level injury and poisoning data for 2000-2012. These variables are now available through the IHIS data extract system.
- Full documentation is available for over 1,000 additional variables. Topics covered by these fully documented variables include: breast exams, child schooling, cancer knowledge, diabetes, firearms, food knowledge, health information sources, radon, and vaccinations.
- he variables HOMELESSEV1 and HOMELESSEV2 were combined into a single variable HOMELESSEV. Although the survey question is the same across 2000-2010, the universe for 2000-2001 is substantially different from the universe for 2002-2010. Data users should review the Comparability tab for HOMELESSEV for full details.
- The variable BREXLIFE was eliminated. The data available from BREXLIFE can be found in a more accessible format in BREXSELFTP, which reports the time period associated with the frequency that the respondent performs self-breast exams, and BREXSELFNO, which reports the number of times the respondent performs self-breast exams per time period.
- The variable HAVEROUTPL was eliminated. The data available from HAVEROUTPL can be found in the harmonized variable USUALPL, which reports if the person has a usual or routine place for medical care.
- The following variables were eliminated from the Cancer Family History variables group: BFCXCAN, FBCXCAN, BSCXCAN, BSOVCAN, FBOVCAN, BFOVCAN, BSUTCAN, FBUTCAN, BFUTCAN, BMPSCAN, BDPSCAN, FSPSCAN, BMTTCAN, BDTTCAN, and FSTTCAN. These variables were included as part of the exhaustive list of cancer variables that report if family members had certain types of cancer; however, the eliminated variables represented unrealistic combinations of sex and cancer type. For example BFCXCAN reported if the person's biological father had uterus cancer. Variables deleted were related to cervix, ovary, prostate, testis, and uterus cancer. No variables related to breast cancer were eliminated. There were no recorded "yes" or "mentioned" observations for any of these eliminated variables.
- A typographical error in the data source for the variable VITBUY was corrected; this will affect VITBUY frequencies for 2012.
- In response to a user query, IHIS staff corrected a recoding error and changed the response categories for DIFTNONEED.
- IHIS has revised the extensive list of food frequency variables to improve comparability across years; to avoid confusion between the newly integrated variables and those that were only available for one or two samples, many food count variables have been eliminated.