Hospital Discharge Data and the Uninsured

275 Brief communication HOSPITAL DISCHARGE DATAAND THE UNINSURED To the editor: Computerized abstracts of patient discharge records are increasingly available to researchers interested in analyzing use of medical care by the uninsured. These abstracts, created soon after the discharge of hospitalized patients, are based on information available in the medical record at the time of discharge. Although the abstracts can form the basis of a promising analytic strategy for monitoring access, their accuracy is uncertain. In one of several studies1"1 of the accuracy of clinical variables contained in discharge abstracts, up to 41.7 percent of surveyed discharges did not meet the clinical criteria for the final diagnosis coded in the hospital discharge abstract.1 Another study showed that insurance claims data failed to identify more than one-half of patients with prognostically important conditions.3 Discharge abstracts reveal only the expected payer of a hospital bill, and not final payment information. Once hospital staff indicate "self-pay" or "no charge" under "expected source of payment," this designation is rarely compared against actual population-based data. Thus there is reason to be concerned about systematic errors in the identification of uninsured patients. For example, previous studies have shown that from 17.3 percent to 72 percent of hospital charges incurred by "self-pay" patients are in fact paid.5"7 This high percentage of hospital charges paid by presumably poor patients suggests that not all "self-pay" patients are indigent and uninsured. Yet health services researchers interested in hospital care for the uninsured have uniformly used a hospital's designation of "self-pay" or "no charge" to identify uninsured and indigent patients.8"14 If the "self-pay" designation is not accurate, then studies that identify uninsured patients only from discharge abstracts may be flawed. To explore the prevalence and magnitude of possible inaccuracies in hospital discharge abstracts, I used detailed financial records from the accounting department of one hospital to verify the accuracy of insurance coding in that hospital's discharge abstracts. The pilot study was performed at an urban community teaching hospital with housestaff outpatient clinics, an emergency room, and a high-risk obstetrical service. As the recipient of federal funds under the Hill-Burton Act, the hospital is required to provide free service to indigent patients. Discharge abstracts were selected from the discharge data set submitJournal of Health Care for the Poor and Underserved · Vol. 5, No. 4 · 1994 276___________________________________________________________ ted to the state reporting agency. All abstracts during two consecutive financial quarters which listed "self-pay" as the expected source of payment were selected. This process yielded records of 472 patients. Financial folders from the hospital accounting department were located for 460 of the 472 patients (97.4 percent). The financial folders showed that 298 (64.8 percent) of the "self-pay" patients received some health insurance benefits and only 162 (35.2 percent) were actually uninsured. Medicaid paid for the hospitalization of 110 patients (23.9 percent); private insurance or Medicare eventually paid at least a portion of the hospital bills of an additional 181 patients (39.3 percent). Conversely, from a random group of 100 patients listed in discharge abstracts as having health insurance coverage, eight (eight percent) were uninsured and received no insurance payment toward the hospitalization. Six were incorrectly listed as having Medicaid coverage, and two were incorrectly listed as having private health insurance. In this hospital, although the insurance coverage listed on discharge abstracts was generally correct, there was a high rate of errors for patients listed as "self-pay." Nearly two-thirds of these patients received health insurance payments for their hospitalization. The process of billing patients identified payment status more accurately than did discharge abstracts, but patients were often billed as long as six months after discharge. Analyses using only information from discharge abstracts would have incorrectly estimated the number and characteristics of patients without insurance coverage. This suggests that data on insurance coverage are subject to error similar to the error rates which have previously been shown for clinical variables. One source of the inaccuracies in the discharge records at this hospital appears to have been coding errors. While the results of this study are based on...