Association between antipsychotic/antidepressant drug treatments and hospital admissions in schizophrenia assessed using a mental health case register

Abstract

Background: The impact of psychotropic drug choice upon admissions for schizophrenia is not well understood.

Aims: To examine the association between antipsychotic/antidepressant use and time in hospital for patients with schizophrenia.

Methods: We conducted an observational study, using 8 years' admission records and electronically generated drug histories from an institution providing secondary mental health care in Cambridgeshire, UK, covering the period 2005-2012 inclusive. Patients with a coded ICD-10 diagnosis of schizophrenia were selected. The primary outcome measure was the time spent as an inpatient in a psychiatric unit. Antipsychotic and antidepressant drugs used by at least 5% of patients overall were examined for associations with admissions. Periods before and after drug commencement were compared for patients having pre-drug admissions, in mirror-image analyses correcting for overall admission rates. Drug use in one 6-month calendar period was used to predict admissions in the next period, across all patients, in a regression analysis accounting for the effects of all other drugs studied and for time.

Results: In mirror-image analyses, sulpiride, aripiprazole, clozapine, and olanzapine were associated with fewer subsequent admission days. In regression analyses, sulpiride, mirtazapine, venlafaxine, and clozapine-aripiprazole and clozapine-amisulpride combinations were associated with fewer subsequent admission days.

Conclusions: Use of these drugs was associated with fewer days in hospital. Causation is not implied and these findings require confirmation by randomized controlled trials.

Figure 1

Illustration of the mirror-image design. For a given drug, the number of admission days for a given patient was calculated for a period before the first recorded use of the drug (M₁, either 1 or 2 years), and a period of identical duration afterwards (M₂). These rates were then corrected for the overall admission rates, for all patients with schizophrenia, during the same M₁ and M₂ periods (see text). Patients were excluded who had no admissions falling in the M₁ period, thus selecting for patients with relatively severe disease. A central gap (C₁+C₂, each 30 days) was excluded to reduce the effects of regression to the mean and errors caused by small inaccuracies in the temporal recording of drugs.

Figure 2

(a) One-year and (b) 2-year mirror-image analyses, showing the change in admission days per patient per year (admission days per year after drug, minus admission days per year before drug, corrected for overall admission rates; see Figure 1 and text). The number of subjects contributing to the measurement for each drug is shown in parentheses (n), with mean age in years (at first use of the drug in question, i.e., in the middle of the period considered) and the number of males (M). Points show means and error bars show 95% CIs; filled symbols indicate that the CI excludes zero.

Figure 3

Association of admission rates during a given 6-month calendar period with drug use in the preceding calendar period (n=1,406). The results are expressed as a change in the number of admission days per patient per year (means±95% CI); filled symbols indicate that the CI excludes zero. Interaction terms are expressed with colon notation. The number of patients taking each drug in at least one time period (whether alone or with another drug) is shown in parentheses (n); for interactions, this is the number of patients who took both drugs during the same time period, for at least one period. The effect sizes were derived from a regression analysis taking account of all other antipsychotic and antidepressant drugs in the analysis (see text), plus 6-month calendar period number (to account for overall trends over time), age, time since the first recorded diagnosis of schizophrenia (in years), and sex.