Study design
This study is a retrospective cohort study of patients with SMI/AP in Dutch family practice.
Study population and Procedure
We followed the STROBE guidelines for reporting observational studies(39). Our data were derived from a de-identified database, the Radboudumc Technology Center Health Data. This database contains Electronic Medical Records (EMRs) of family practices with information on patient demographics, diagnoses and symptoms, laboratory test results and drug prescriptions, number of visits (i.e. visits to the practice) along with characteristics of the family practices such as number of patients registered and geographical location. Drug prescriptions are coded according to the WHO Anatomical Therapeutic Chemical (ATC) Classification system.(40) Diagnoses and symptoms are coded according to the International Classification of Primary Care (ICPC)(41). The database provides reliable data because in the Netherlands nearly all people are registered in a family practice over a long period of time, and FPs are used to classify each visit, using the ICPC system. The FP operates as a “gatekeeper” for secondary care and consequently medical specialists inform the FP about diagnosis and treatment.(42) However, electronic records for outpatient psychiatric visits in the Netherlands are separate from the FP’s system. Therefore, visits to a psychiatrist and data concerning CVR collected there were not included. We selected patients who have an indication for yearly assessment of CVR based on their psychiatric disorder or based on the use of antipsychotic medication or lithium.
We used data from 151.238 persons, who were listed in any of the 24 involved family practices, selected by region and availability of data from our FP database, between January 2013 and December 2014. We selected patients with (I) schizophrenia, affective psychosis, bipolar disorder or psychosis not otherwise specified (NOS) with a diagnose date prior to 1-1-2013 or (II) at least two prescriptions of antipsychotics, or (III) a prescription of lithium, II and III prescribed for the first time before 1-7-2013. This date was chosen since we only had access to the prescription records in this defined study period. Patients were excluded if (I) aged younger than 18 years, (II) diagnosed with dementia, (III) diagnosed with delirium without the presence of a psychotic disorder, (IV) if they were not registered for more than 12 months in the selected family practice in our study period, since FPs usually assess a patient’s CVR profile once a year(43) and (V) diagnosed with rheumatoid arthritis, since CVR assessment in this patient category was introduced just before our study period and therefore could possibly confound our results.(43, 44)
Data collection
Patients with SMI/AP were divided into three groups (I) patients without another indication for yearly assessment of CVR according to the current FP guidelines(43) ‘SMI/AP-only group’. (II) Patients with SMI/AP and diabetes mellitus (DM), and thus an extra indication for CVR assessment ‘SMI/AP+DM group’. (III) Patients with SMI/AP and a history of a cardiovascular disease (CVD; i.e. stroke, angina pectoris, acute myocardial infarction, transient ischemic attack, intermittent claudication and aortic aneurysm), and therefore an extra indication for CVR assessments ‘SMI/AP+CVD group’. Patients with both DM and CVD at baseline were added to the SMI/AP+DM group because patients with DM are routinely part of a chronic care program that pro-actively invites patients for monitoring.
Our primary outcome measure was the screening rate of CVR, i.e. the proportion of patients in each subgroup that received screening for their CVR factors in the defined study period.
The CVR factors were selected as recommended in the Dutch FP guidelines (i.e. Body Mass Index (BMI), blood pressure, estimated Glomerular Filtration Rate (eGFR), smoking status, fasting glucose, lipid spectrum, use of alcohol, family history of cardiovascular disease).(43) However, considering the observational nature of this study and screening criteria described in previous studies,(29, 30) we included a broader range of assessments (Appendix A1).
We divided the observed screening in three levels: adequate, moderate and insufficient, based on current Dutch FP guidelines.(43) The screening rate was considered ‘adequate’ when BMI, smoking status, blood pressure, glucose and cholesterol/HDL ratio were all recorded at least once during the observation period, since these are the assessments that are needed to assess the 10-year CVR of a patient and provides the indications for cardiovascular risk-lowering medication. The screening rate was considered ‘moderate’ when the assessment included BMI, smoking status and blood pressure, as these can be measured without a blood test. The screening rate was considered insufficient if it did not meet up to these requirements. A 2-year window was chosen to gain insight in the role and awareness of the FP in this matter. Since FPs usually invite their high-risk patients once a year, patients who were screened just over the 1-year time window because of a delay in their response, would be part of the unscreened group, which would underestimate the screening rate.
Moreover, we wanted to identify factors associated with any CVR screening (adequate or moderate). The following factors were studied: age, sex, type of psychiatric disease, use of antipsychotics, use of antidepressants, CVR medication (i.e. statins, blood pressure drugs and aspirin), COPD, abuse of alcohol or drugs, any records of social issues and frequency of visits. We selected ICPC-codes concerning diseases and social problems (see Appendix A2) and prescription records of antidepressants for this purpose. The ATC codes of AP, lithium and antidepressants are listed in Appendix A3. We also selected (home) visits and calculated the frequency of visits per year of each patient.
Statistical analyses
Descriptive analyses were used to describe the patient characteristics and to provide insight in the screening rate in the three different patient groups. As a result of the hierarchical structure of the study (patients nested within practices), multilevel analyses (random intercept model) were performed that took into account the variability associated with each level of clustering. Logistic regression analysis was performed to test the differences in screening rates between the three groups .
In addition, for the SMI/AP-only group we investigated the patient characteristics from table 1 that were associated with an adequate or moderate screening rate. First, we included characteristics for the multivariate model that were univariately associated with screening ( p < 0.20). After that a backward regression analyses was performed with these characteristics . A p-value of < 0.05 was considered to be statistically significant, based on two-sided tests. A sub analysis was added to show if the results differ between two groups: patients who were included based on their diagnosis (SMI) and patients who use AP without a diagnosis that suits the use (addendum). All analyses were carried out using IBM SPSS statistics 22.0.