How Good Are Hospital Data from a Household Survey?

Abstract

The study is based on hospitalization data from a morbidity survey of 1100 families (2950 individuals) comprising 50% of the dwellings in Kit Carson County, Colorado. During the 12-month period preceding the survey, 31 widely scattered hospitals were used by these families. Complete rosters of admissions were obtained from the 2 hospitals in the County. Only 93% of the hospitalized cases reported in the survey could be matched with hospital rosters. A number of individuals included in the survey, who according to the hospital records were hospitalized in the 12-month period, were not reported hospitalized in the survey; a somewhat smaller number, who were reported to have been hospitalized in the 2 hospitals in the County, were not shown on the hospital rosters. Comparison of days of hospital care reported in the survey with that obtained from the hospital records completely agreed in only 52% of the cases. These accounted for 41% of the total days hospitalized according to the hospitals. But, because of compensatory errors, the over-all number of days reported by the survey exceeded the days shown by the hospital by about 8%. Information on days of hospitalization for those reporting hospitalization in the survey was available from 2 additional hospitals outside of the County. Agreement between days of hospitalization per admission as given in the survey, with the hospital records from these 4 hospitals, holds true for about 44% of the cases; accounting for less than 36% of the days reported by the hospitals. Because of compensatory errors, the total days derived from the survey actually exceed by about 8% the days derived from the 4 hospital records. The study shows that a host of factors including place of residence of patient, his age, sex, recency of hospitalization, duration of hospitalization, relationship to the respondent, etc., affect the level of agreement between the information obtained in the survey, and the hospital record.