User Note - Insurance Data Collection

Changes in the Insurance Data Collection Process

The order and wording of NHIS questions on health insurance coverage changed over time to reflect changes in the availability of different types of coverage and issues of interest.

Prior to 1972, respondents were asked whether family members had hospital and doctor/surgical coverage only. From 1972 to 1996, a different approach was used: respondents were asked a series of questions about whether family members had each of several types of coverage (e.g., Medicare, Medicaid, VA). Starting in 1997, respondents selected the type(s) of coverage held by family members from a list of different types of coverage. This list of different kinds of coverage changed over time in response to the availability of public programs (e.g., adding Children's Health Insurance Program beginning in 1999). For all years prior to 2019, respondents provided information on health insurance for all co-resident members of their family. For 2019 forward, as part of the NHIS redesign, health insurance information was only collected for the sample adult and sample child. For more information on the redesign, see the user note.

I. 1970

Interviewers began by asking whether the sample person (for persons 65 or over) was covered by the part of Medicare that pays for hospital bills (Part A) and about coverage through the part of Medicare that pays for doctor's bills (Part B). They then asked to see the Medicare card to verify coverage. The interviewer next asked whether anyone in the family was covered by private insurance, defined as insurance that paid any part of a hospital, doctor's or surgeon's bill (excluding plans that pay only for accidents). Respondents were first asked whether anyone in the family had hospital insurance, and then were asked about insurance for any part of a doctor's or surgeon's bill.

Back to Top

II. 1976

Interviewers began by asking whether the sample person (for persons 65 or over) was covered by any part of Medicare. (No definition of Medicare was provided to respondents, and no definition was provided to interviewers in the Field Representative's Manual.) Interviewers then asked whether, in the past 12 months, anyone in the family had received Medicaid (defined as a public assistance program that provides medical assistance to persons in need). If the answer was "yes," the interviewer asked to see the Medicaid card to verify coverage. Next, the interviewer asked whether anyone in the family was covered by private insurance (though the term "private" insurance was not used in survey questions until 1993). As an operational definition of private insurance, the interviewer referred to all kinds of insurance except those that pay only for accidents (and excluding Medicare and Medicaid). Respondents were first asked whether anyone in the family had hospital insurance, and then were asked about insurance for any part of a doctor's or surgeon's bill.

Back to Top

III. Using Private Insurance Plan Information for 1980 to 1996

For 1980-1996, as in other years, respondents provided information about the health insurance coverage of all their co-resident family members. For 1980 to 1989, information was collected on up to five plans (six plans in 1983). For 1992 to 1996, information was collected on up to four plans. For 1980 to 1996,the details of private plans held by any family member (such as whether the plan was obtained through work, in HI1WORK) were assigned to each family member's record, regardless of whether the individual was covered by the plan. To identify persons covered by each plan, the variables HI1PCOV, HI2PCOV, HI3PCOV, HI4PCOV, HI5PCOV (1980 to 1989) must be used, to clarify whether the person is actually covered by any of the health insurance plans on his or her record. For example, for 1996, to determine whether a person is covered by private health insurance obtained through the workplace, analysts must look for affirmative responses to HI1PCOV and HI1WORK, HI2PCOV and HI2WORK, HI3PCOV and HI3WORK, and HI4PCOV and HI4WORK.

Because of the complexity involved in determining an individual's coverage status in these years, the IPUMS staff created recoded variables, such as HIPWORKR, which include only those covered by plans and combine responses across multiple plans.

Back to Top

IV. Data Collection in 1978, 1980, 1982, 1984, 1986, 1989

The interviewer began by reading a definition of Medicare: "Medicare is a Social Security health insurance program for disabled persons and for persons 65 years old and older." The interviewer asked whether anyone in the family had Medicare and, if so, whether the person had Medicare for hospital bills (Part A) and/or Medicare for doctor's bills (Part B). To verify coverage, the interviewer asked to see the Medicare card.

The interviewer next asked about private insurance (though the term "private" was not used until 1993). Private insurance was operationally defined to respondents as all kinds of insurance except those paying only for accidents (and excluding Medicare). Respondents were asked in turn about hospital insurance and about insurance for any part of a doctor's or surgeon's bill. For 1986 and 1989, the end of the question also referred to insurance for dental bills.

Back to Top

V. Question Wording in 1986 and 1989 on Dental Coverage and Single Service Plans

In 1986, interviewers asked, "(Not counting Medicare) Is anyone in the family covered by a health insurance plan which pays any part of a hospital, doctor's or surgeon's bill?" They then asked, "Is anyone in the family now covered by any other health insurance plan which pays any part of a hospital, doctor's, surgeon's or dentist's bill?"

In 1989, interviewers instructed respondents not to include plans that pay for only one type of service. Interviewers asked, "(Not counting Medicare) is anyone in the family now covered by a health insurance plan which pays any part of hospital, doctor, or dental bills?" They then asked, "Is anyone in the family now covered by any OTHER health insurance plan? Again, do not include plans that pay for only one service." Note that the reference period for coverage was the preceding month, and that that single service plans were not considered to be health insurance.

Back to Top

VI. Data Collection in 1993-1996

In 1993 to 1996, interviewers began the insurance inquiry with the statement, "There are several government programs that provide medical care or help pay medical bills." Interviewers then asked whether anyone in the family received Medicaid and, if necessary, supplied the following definition: "Medicaid or (local name) is a public assistance program that pays for medical care." Similar question wording was repeated for other government insurance programs, including military health care, CHAMPUS, CHAMPVA, or VA coverage, and any public assistance programs, other than Medicaid, that paid for health care.

To collect information about private coverage, interviewers asked, "(Not counting the government health programs we just mentioned), In (month) was anyone in the family covered by a private health insurance plan?" Again, a definition was read, as necessary: "Besides government programs, people also get health insurance through their job or union, through other private groups, or directly from an insurance company. A variety of types of plans are available, including health maintenance organizations (HMOs)."

Note that single service plans were considered to be health insurance during 1993-1996 (which was not true in other periods).

Back to Top

VII. Data Collection in 1997-2018

For 1997 forward, interviewers first asked, "Are you covered by health insurance or some other kind of health care plan?" and then queried, "What kind of health insurance or health care coverage do you have?" For 1997 to 1999, only those who answered the initial question affirmatively were asked follow-up questions on the kind of health insurance and the details of insurance coverage. By contrast, beginning in 2000, those without an affirmative response to the initial question were still considered "in the universe" for subsequent questions and have data for follow-up insurance questions.

For 1997-2018, after the opening question, respondents were handed a flashcard listing various insurance types and asked to mark all that applied (see list below). In 1997-1998, respondents were instructed to exclude plans that pay for only one type of service (e.g., nursing home care, accidents, or dental care). Beginning in 1999, this policy changed. The category "Single Service Plan (e.g., dental, vision, prescriptions)" was added to the flashcard, and respondents could select that category.

Back to Top

VIII. Data Collection in 2019 forward

Starting in 2019 health insurance information was only collected for the sample adult and sample child, not all co-resident members of their family as was the case in years prior to 2019. For 2019 forward, after the opening question, interviewers read each of the response options. Single Service Plans for dental, vision, and prescription coverage were asked as three separate questions during this period and were not included in the list of categories that respondents could select as possible sources of coverage.

Back to Top

IX. Insurance Coverage Categories for 1997 forward

The various types of insurance coverage that persons could indicate from 1997 forward included:

In follow-up questions, interviewers recorded the names of up to four private health insurance plans for 1997-2018, and two private health insurance plans for 2019 forward. If the person was reported as covered by a public program such as CHIP, a state-sponsored health plan, or by another public program (other than Medicaid) that paid for health care, the interviewer recorded the name of that plan. The placement and wording of the questions about the names of specific government health care plans varied across years.

For 2004-2018, information is publicly available only on the first and second private insurance plans. However, individuals with more than two private plans are rare, as is shown by analysis of HIPRIVGT2, which identifies persons with three or more private plans.

Back to Top

X. Data Editing and Recoding

During the course of data editing for 1997 forward, the NCHS discovered many errors in the responses to questions about insurance coverage. Often, respondents misclassified the type of insurance they had. This might be shown, for example, by a mismatch between the verbatim name of an insurance plan and the type of insurance coverage the person selected from the categories available.

Accordingly, the NCHS created a series of recoded insurance variables. For these recoded insurance variables the data are back-edited, taking into account such factors as the proper classification of the verbatim names of insurance plans and responses to questions about why insurance coverage had stopped. The NCHS strongly advises analysts to use these recoded insurance variables as a more reliable source of information about the types of insurance coverage than is provided by respondents' original and unedited answers about their insurance type. Back-edited insurance variables include:

Back to Top