Appendix A ICD-9 codes used in the study

Codes for pre-eclampsia and related disorders

642.4 Mild pre-eclampsia

642.5 Severe pre-eclampsia

642.6 Eclampsia

642.7 Toxaemia

Codes for control groups

664.0,6641 1st or 2nd degree perineal laceration

650.0 Normal delivery

644.0 Threatened preterm labour

656.3 Fetal distress

654.2 Caesarean section in previous pregnancy

661.2 Uterine inertia

669.5 Forceps delivery

658.1 Premature rupture of membranes

645.0 Prolonged pregnancy

663.3 Entangled cord

Codes for venous thromboembolic disorders

451.1 Deep vein thrombosis

415.1 Pulmonary embolism and infarction

Appendix B Accuracy of ICD-9 codes for pre-eclampsia

We tested the accuracy of pre-eclampsia coding at one hospital by reviewing the records of a random sample of 197 patients. Of these, 99 had ICD-9 codes for pre-eclampsia (appendix A) as the primary diagnosis. The other 98 patients had codes for one of the control groups (appendix A). The chart reviewer was blinded to the primary diagnostic codes. To be classified with pre-eclampsia, patients had to meet accepted criteria for pre-eclampsia: hypertension (blood pressure of 140/90 mm Hg, or 15/30 mm Hg above baseline, measured on two occasions six hours apart) and proteinurea (1+ protein on dipstick, >300 mg in a 24 hour urine collection, or documented proteinurea before admission). Overall, 62 people (31.5%) satisfied the criteria for pre-eclampsia. This reabstraction study found that the pre-eclampsia codes had a sensitivity of 89% (95% confidence interval 78% to 94%) and a specificity of 67% (79% to 94%) for patients with true pre-eclampisa.

Appendix C Accuracy of ICD-9 codes for venous thromboembolic disorders

Admissions for thromboembolic disorders were determined from the discharge abstract database using the codes listed in appendix A. We tested the accuracy of these codes at one hospital by reviewing a random sample of medical records of 135 patients. Of these, 72 had ICD-9 codes for thromboembolic disorders (415.1 and 451.1) as the primary diagnosis, secondary diagnosis, or a complication. The remainder had control conditions that included pneumonia (486.X, 482.3), congestive heart failure (428.X), venous compression (459.2), or other disorders of the circulatory system (459.8). The chart reviewer was blinded to the final codes in the discharge abstract. Patients were classified with a thromboembolic admission if they had a positive pulmonary angiogram, a high probability ventilation-perfusion scan, or a compression ultrasound or venogram documenting an obstructive clot in the deep venous system of the leg or arm. The patient was also classified with a thromboembolic admission if the treating physician stated in the chart that the patient had a venous thromboembolic event and the patient was discharged on anticoagulants. The reabstraction study found that these codes had a sensitivity of 97% (95% confidence interval 89% to 99%) and a specificity of 90% (79% to 94%) for patients with true thromboembolic disease.