patient reason present in a majority of the poisoning sample, that is, in each case where an unrelated ICD-9 code was also present.
A total of 11,533 records with an ICD-9-CM diagnosis or E-code of poisoning was extracted from NHDS 1997–2001 files, representing an estimated total of 291,000 annual hospitalizations, or 0.8 percent of all estimated discharges (Table 3-8). Poisoning-related hospitalizations were more likely for females than males (120 per 100,000 compared with 90 per 100,000). Rates of hospitalization increased with age.
Because 25 percent of all discharges were associated with “unknown” race/ethnicity, even though the available data for patient race imply that whites are less likely to be hospitalized for poisoning than blacks, this observation must be viewed with caution. Hospitalization visit rates did not demonstrate substantive variation by region. Discharge status information was available for most of the visits, indicating that the majority of poisoning cases are discharged home. An estimated 9 and 6 percent of hospital visits were discharged to other short- and long-term care facilities, respectively. Because the short-term care facilities included in the discharge status variable could include some (but not all) health care facilities not actually incorporated into the NHDS sampling frame (long-term care facilities as a category are excluded), 9 percent is too high an estimate of multiple hospitalizations per episode to be used as a discounting rate to convert these poisoning hospitalizations into episodes.
A source of admission variable was added to NHDS in 2001. Examination of all hospitalizations by the source of admission (Table 3-9) indicates that 65 percent of all cases (186,000 visits) were admitted from an emergency department, followed by 11 percent (31,000) from physician referral and 2 percent (5,000) from another hospital. Because 19 percent of all hospitalizations in the file were missing the source of admission information, the emergency department, physician, and hospital transfer sources of admission may actually be higher. Thus a figure of 3 percent for hospital transfers is used for later incidence estimates, which would represent the difference between episodes and true incidence. This presumes, however, that readmission of an individual for the same poisoning episode is extremely infrequent.
Two-thirds of the poisoning inpatient visits had a poisoning ICD-9 code as the principal diagnosis on the discharge abstract, with the remainder as a secondary listing (e.g., a primary diagnosis of aspiration pneumonia in a concomitant drug overdose). The six most common ICD-9-CM codes listed were “suicide and self-inflicted poisoning by tranquilizers” (E950.3), “poisoning by benzodiazepine-based tranquilizers” (969.4), “poi-