For 1969-1981, nursing homes and convalescent homes were also considered hospitals, in the context of this question. In those years, separate verification questions for stays in nursing homes or convalescent homes were included, to ensure these stays were counted (so long as the person who had stayed in such an institution was a household member at the time of the interview). Hospitalizations for deliveries were also included, and interviewers probed to ensure that deliveries counted as (overnight) hospitalizations for both the mother and the baby.

The Field Representative's Manuals for 1969-1996 state that, for statistical purposes, the actual reference period for the hospitalizations reported in HOSPNGHTD is 13 to 14 months prior to the interview. For example, as the Manual for 1982 explains:

The Field Representative's Manuals for 1969-1970 noted.

Similarly, the Manual for 1972 defined being a "patient" as being admitted and staying overnight or longer in a hospital, excluding the emergency room or outpatient clinic. (This definition was not routinely shared with respondents.) From 1982 forward, the Field Representative's Manuals also instructed interviewers to exclude "stays" in the hospital for nonmedical reasons, such as a parent staying with a sick child. Interviewers were directed to "not include stays in the hospital during which the person does not spend at least one night, even though surgery may have been performed," and also to exclude cases in which the person was admitted to and released from the hospital on the same date.

Between 1969 and 1996, respondents were asked to report the exact dates, or their best estimate of the dates, for each hospitalization. Interviewers were instructed to probe to determine the dates as closely as possible. The Manuals urged, "Use the calendars and the list of holidays in your Flashcard Booklet to assist the respondent in recalling dates," and provided examples of appropriate probe questions (e.g., "Was it before or after Memorial Day (or some other holiday)?")

Considerable space was devoted to this issue, with the Manuals concluding:

If adequate information about the date of the hospitalization was not provided (if, for example, the spaces for filling in month, date, and year of the hospitalization on the questionnaire were left blank), then no hospital record was generated for that stay was generated by NCHS.

For 1982, the survey text obliquely included nursing homes in the wording, "How many nights was -- in the hospital (nursing home)?," but the earlier follow-up questions about "Was anyone in the family in a nursing home, convalescent home, or similar place since [date] a year ago?" disappeared. From 1983 forward, the use of the parenthetical "nursing home" was removed from the survey question text. The effect of this change may be modest, for a note in the NCHS 1982 Current Estimates states, "Prior to 1982, few additional hospitalizations were in fact detected by this question." The shift may, however, result in a slight drop in the number of overnight hospital stays reported for older people after 1982.

For example, only the question wording for 1997 forward explicitly included the directive, "Do not include an overnight stay in the emergency room." Also, beginning in 1997, interviewers did not probe to include deliveries (although the Field Representatives Manuals prompted interviewers to "include new mothers and babies who were hospitalized for the baby's birth.") Further, HOSPNGHT is based directly on answers from respondents; HOSPNGHTD results derive from hospital records generated at NCHS, with the source data having to pass certain quality checks, such as adequate information on the date of the hospitalization. For these reasons, researchers should exercise caution in comparing pre-1997 results from HOSPNGHTD with post-1997 results from HOSPNGHT.

As the 1997 NHIS Survey Description, states: