258 



For those ICD-9 codes which explicitly mention alcohol or drug abuse in their titles, 

 we assigned a PAR of 100%, since all of these hospital days are attributable to substance 

 abuse. In addition, the NIAAA has identified a list of diagnoses that are completely alcohol- 

 related (e.g. cirrhosis). These diagnoses were also assigned a PAR of 100% (Table 5). 



The costs to Medicare of substance abuse treatment in psychiatric hospitals was also 

 included in the study. This was derived from data collected by the National Association of 

 Psychiatric Health Systems on both the use of drug and alcohol services and the prevalence of 

 Medicare discharges.'" 



Database 



To determine the Medicare hospital costs for treating substance abuse-related 

 illnesses, we used Medicare data repxDrted on the 1991 National Hospital Discharge Survey 

 (NHDS). The NHDS is a nationwide sample survey of short-stay hospitals. Each NHDS 

 record includes the patient's primary payer, demographic information, principal diagnosis and 

 up to four secondary diagnoses (reported by ICD-9 codes), DRG category, procedures, and 

 length of stay. 



Extracting all Medicare discharges that had a primary diagnosis that fell within a 

 given ICD-9 code for which we had a PAR, we then applied each PAR (by age or sex, if 

 applicable) to the discharges with corresponding diagnoses. For example, approximately 

 1 1 1 ,000 Medicare hospital discharges had lung cancer as their primary diagnosis. Of these, 

 87% (the PAR) or 96,600 hospitalizations were attributed to smoking. 



Since Medicare pays on the basis of DRGs (not diagnoses), it was necessary to 

 analyze the data by DRGs to estimate the costs of these substance abuse-related admissions to 



15 



