(151) Use of Patient Profiles to Validate Acute Myocardial Infarction in Claims Data

Publication Number: 1196

Background: Chart review – which is the gold standard for outcome validation – is not permissible or feasible in some claims databases. Claims databases, however, contain a wealth of information that can be used to assess the validity of outcomes.

Objectives: To assess the validity of acute myocardial infarction (AMI) using patient profiles extracted from pseudonymized German claims data.

Methods: All hospitalizations with a main diagnosis of AMI (ICD-10-GM codes I.21 and I.22) in the German Pharmacoepidemiological Research Database (GePaRD) between January 2016 and December 2017 were eligible, if the respective patients were (i) ≥50 years, (ii) continuously insured ≥1 year, had (iii) valid information on sex, and (iv) residency in Germany. Of those, 250 cases were randomly selected (50% males) and de-identified patient profiles were built based on (i) demographic information and including information on timing relative to the diagnosis of AMI: (ii) all hospitalizations, all diagnoses and procedures, and discharge reasons, (iii) all outpatient diagnoses including diagnostic certainty and specialty of the diagnosing physician from up to 5 years before and up to 1 year after the event and (iv) all outpatient encounters and (v) all outpatient dispensings up to 1 year before and up to 1 year after the event. Using adjudication criteria based on clinical guidelines and risk factors, 2 trained physicians independently classified cases as ‘definite’, ‘probable’, ‘unlikely’, and ‘non-assessable’. Positive predictive values (PPVs) were calculated as the percentage of confirmed cases (‘definite’ or ‘probable’) among all cases excluding those classified as “non-assessable”.

Results: 65,298 AMI cases were eligible for validation (22,804 females, 42,494 males). Agreement between experts was good, 93.6% of cases were classified concordantly: 228 as ‘definite/probable’, 4 as ‘unlikely’ and 2 as ‘non-assessable’. After discussion of the remaining 16 profiles with a third physician, 242 cases were classified as ‘definite/probable’, 6 as ‘unlikely’, and 2 as ‘non-assessable’. This resulted in a PPV of 97.6%. In the 125 men, 123 cases were classified as ‘definite/probable’, 1 as ‘unlikely’ and 1 as ‘non-assessable’, yielding a PPV of 99.2%. In the 125 women, 119 cases were classified as ‘definite/probable’, 5 as ‘unlikely’ and 1 as ‘non-assessable’, yielding a PPV of 96.0%.

Conclusions: German claims data contain information that allows to “retrace” patient histories for outcome validation. Overall, physicians were satisfied with the available level of detail. Validity of AMI identified by hospital diagnoses is high in German claims data.