Effects of age on length of hospital stay in a low-income population.

A major focus of efforts to monitor and control health care expenditures has centered on factors related to hospital length of stay. Age, in particular, is usually assumed to be an important correlate of hospital length of stay, especially when diagnosis and hospital variables are also taken into account. In an analysis of data representative of Medicaid cases in 23 large New York City hospitals during 1972, these effects turn out to be less clear-cut than generally assumed. Although the first-order coefficients of length of stay regressed on age and the rank order correlation of the means are statistically significant, such wide differences in length of stay within age groups remain as to reduce the substantive impact of the observed findings. This same pattern of wide variations within age groups persists even when diagnosis is controlled by use of up to 80 covariant groups. In over half the 80 diagnostic categories, no significant age effects was found. When the age analysis is repeated using 23 hospitals as the covariants, estimated age effects differed between institutions, and a similar pattern of large within-group variation was observed. These findings are interpreted as a caveat to health care researchers who might otherwise plan overly sophisticated utilization review systems.