It was conjectured that a small group of chronically ill in tertiary ambulatory clinics consume a large amount of health resources and that, from the perspective of the patient, psychosocial rather than disease variables would most explain their health service utilization and subsequent cost. New referrals with a chronic illness (N = 215) to one of three clinics (oncology, rheumatology, and gastroenterology) consented to participate in a subsequent trial of a psychosocial intervention designed to promote their adjustment to illness and, conceivably, to reduce their health service utilization. At baseline an inventory to describe the disease, treatment, functional capacity, prognosis, and socioeconomic situation of consenting subjects was completed. In addition, subjects completed the Psychosocial Adjustment to Illness Inventory (PAIS-SR), the Family Assessment Device (FAD), the Meaning of Illness Questionnaire (MIQ), and a Health Service Utilization Inventory designed to assess direct and indirect costs of health resources. These data were entered into a concurrent analytic survey design. Participating subjects represented a more socioeconomically advantaged and better-adjusted group of chronically ill patients compared with others referred to the tertiary clinics. They were representative of all new referrals in their use of the majority of health services. However, once hospitalized, participating subjects stayed longer and used specialists less. There was no important relationship between disease severity or prognosis and any type of service utilization, including hospitalization. Because the strongest correlate of all types of health services consumed was psychosocial adjustment to illness (r = 0.28 to 0.33), patients were partitioned into one of three categories of adjustment to illness: good, fair, and poor. The total annual cost per patient was $23,883, if poorly adjusted, compared with $9,791 if well adjusted. If cash transfers (benefits paid by different types of insurance) are added, the average 1987 annual cost per poorly adjusted subject was $31,291 per patient, compared with $13,771 for a patient well adjusted to the illness. There was a statistically significant and economically important linear gradient in 1987 treatment costs per category of adjustment. The possible economic implications for psychosocial intervention are highlighted.