Diagnostic codes for sudden cardiac death and ventricular arrhythmia functioned poorly to identify outpatient events in EPIC's General Practice Research Database.

PURPOSE: To examine the validity of OXMIS/Read diagnoses of hospitalization for sudden cardiac death and ventricular arrhythmia (SD/VA) for use in studies of arrhythmogenic effects of outpatient medications in the General Practice Research Database (GPRD). METHODS: We identified putative occurrences of hospitalization for SD/VA in patients receiving outpatient prescriptions for cisapride, domperidone, or metoclopramide. We then administered a questionnaire to general practitioners (GPs) caring for these patients, requested hospital discharge consult letters, and examined the positive predictive value (PPV) of diagnostic codes in identifying outpatient events precipitating hospitalization. RESULTS: We identified 84 putative events, 38 (45%) of which occurred in patients cared for by GPs participating in the follow-up scheme. Thirty of 38 questionnaires (79%) were completed. The PPV for the occurrence of any SD/VA was 93% (95% CI, 78-99%). However, the PPV for SD/VA occurrence in an outpatient setting precipitating a hospitalization was only 23% (95% CI, 10-42%). The majority of inpatient diagnostic codes reflected events occurring after hospital admission, not precipitating it. CONCLUSIONS: While computerized codes for SD/VA agreed well with physician diagnosis, they operated poorly to identify events occurring in an outpatient setting. Studies of SD/VA in the GPRD should verify events on a case-by-case basis.

Instantaneous preference was a stronger instrumental variable than 3- and 6-month prescribing preference for NSAIDs.

OBJECTIVE: Prescriber preference has been used as an instrumental variable (IV) in a prior study of nonselective nonsteroidal anti-inflammatory drugs (NSAIDs) vs. selective cyclooxygenase-2 (COX-2) inhibitors, with preference expressed as the drug constituting the immediately preceding prescription by the same prescriber (instantaneous preference). We sought to compare the correlations between different IV measures with exposure. STUDY DESIGN AND SETTING: In an ambulatory electronic medical record database of university-based physicians, we compared correlations with exposure among three measures of prescriber preference: instantaneous preference, and the proportion of that prescriber's prescriptions in the past 3 and 6 months that were for an NSAID. RESULTS: We identified 37,934 initial NSAID/COX-2 prescriptions. The correlation with exposure was 0.283 (95% confidence interval 0.274-0.292) for instantaneous preference, 0.197 (0.187-0.206) for 3-month preference, and 0.170 (0.160-0.180) for 6-month preference. CONCLUSION: Instantaneous NSAID/COX-2 prescribing preference was most strongly correlated, and therefore the strongest IV. Future research should focus on the robustness of IV methods to violations of underlying assumptions, extension of IV methods to more than two groups, ratio measures of association, second and subsequent prescriptions per person, and time-varying exposures.

Quality of Medicaid and Medicare data obtained through Centers for Medicare and Medicaid Services (CMS).

BACKGROUND: The Centers for Medicare and Medicaid Services (CMS) now make data from the Medicaid and Medicare programs available for research. Prior research has identified important issues in Medicaid data obtained from a commercial vendor. OBJECTIVES: To perform exploratory analyses of Medicaid and Medicare data obtained through CMS. RESEARCH DESIGN: We obtained data from 1999 to 2000 for 5 large Medicaid programs (California, Florida, New York, Ohio, and Pennsylvania), together with the corresponding Medicare data for dual eligibles. We first plotted the number of prescription claims per month. We next plotted the frequency of hospitalization by age group, first using Medicaid data alone, then supplemented with Medicare data. We next plotted the frequency of claims with a diagnosis of complications of pregnancy and childbirth, and of prostate cancers, stratified by sex. RESULTS: The number of prescription claims per month was very stable. Because of cocoverage by Medicare, Medicaid data seemed to miss a substantial number of hospitalizations in those age 45 years and older. Diagnoses of complications of pregnancy and childbirth were uncommon in men and in women older than 60 years, and prostate cancer diagnoses were uncommon in women. CONCLUSIONS: In contrast to Medicaid data previously obtained from a commercial vendor, we found no evidence that prescription Medicaid data from CMS were incomplete. We were also reassured by the infrequency of gross diagnostic miscoding. Researchers using Medicaid data to study hospital outcomes should obtain supplemental Medicare data on dual eligibles for studies of persons aged 45 years and older.

Observed association between antidepressant use and pneumonia risk was confounded by comorbidity measures.

OBJECTIVE: A prior study suggested that antidepressants might increase the risk of hospitalization for pneumonia in the elderly. This study sought to confirm or refute this hypothesis. STUDY DESIGN AND SETTING: Case-control study of persons aged 65 and above nested in the UK General Practice Research Database. RESULTS: We identified 12,044 cases of the hospitalization for pneumonia (the primary outcome) and 48,176 controls. The odds ratio (OR) for any antidepressant use, adjusting for age, sex, and calendar year was 1.61 (95% confidence interval 1.46-1.78). After further adjustment for comorbidity measures, the OR was 0.89 (0.79-1.00). We also identified 159 cases of hospitalization for aspiration pneumonia (the secondary outcome) and 636 controls. The OR for any antidepressant use, adjusted for age, sex, and calendar year was 1.45 (0.65-3.24). After further adjustment for comorbidity measures, the OR was 0.63 (0.23-1.71). CONCLUSION: These findings refute the prior hypothesis that use of antidepressants by elderly patients increases the risk of hospitalization for pneumonia or for aspiration pneumonia. Decisions regarding use of antidepressants in elderly persons should not be affected by concern about pneumonia risk. Data-derived hypotheses should be independently confirmed before being acted upon.