A Systematic Review and Meta-analysis of Recovery in Schizophrenia

Erika Jääskeläinen; Pauliina Juola; Noora Hirvonen; John J. McGrath; Sukanta Saha; Matti Isohanni; Juha Veijola; Jouko Miettunen


Schizophr Bull. 2013;39(6):1296-1306. 

In This Article

Abstract and Introduction


Objective: Our primary aims were (a) to identify the proportion of individuals with schizophrenia and related psychoses who met recovery criteria based on both clinical and social domains and (b) to examine if recovery was associated with factors such as gender, economic index of sites, and selected design features of the study. We also examined if the proportions who met our definition of recovery had changed over time.

Method: A comprehensive search strategy was used to identify potential studies, and data were extracted for those that met inclusion criteria. The proportion who met our recovery criteria (improvements in both clinical and social domains and evidence that improvements in at least 1 of these 2 domains had persisted for at least 2 years) was extracted from each study. Meta-regression techniques were used to explore the association between the recovery proportions and the selected variables.

Results: We identified 50 studies with data suitable for inclusion. The median proportion (25%–75% quantiles) who met our recovery criteria was 13.5% (8.1%–20.0%). Studies from sites in countries with poorer economic status had higher recovery proportions. However, there were no statistically significant differences when the estimates were stratified according to sex, midpoint of intake period, strictness of the diagnostic criteria, duration of follow-up, or other design features.

Conclusions: Based on the best available data, approximately, 1 in 7 individuals with schizophrenia met our criteria for recovery. Despite major changes in treatment options in recent decades, the proportion of recovered cases has not increased.


It is widely accepted that a proportion of individuals who develop schizophrenia have a favorable prognosis. Symptoms can abate over time, and a proportion of those with schizophrenia attain good outcomes on a range of clinical and functional outcomes (eg, education, employment, and relationships). The precise proportion of cases that have favorable outcomes is less clearly understood. To a large degree, this relates to uncertainty about how to measure multifaceted outcomes such as "recovery." Considering how much research attention has been allocated to exploring the onset of psychosis (eg, prodrome and early psychosis), it is appropriate that a comparable degree of research scrutiny also be accorded to the recovery of psychosis.[1] With respect to the remission of clinical symptoms, operationalizable criteria are now available.[2–4] However, symptom profiles are only one component of the many facets of recovery. Many consumer-based groups conceptualize recovery as a personal journey (ie, a subjectively evaluated process dealing with symptoms over time) rather than a defined point outcome (completely recovered vs persistent illness).[5] In contrast to most clinical symptoms, outcomes related to recovery do not lend themselves to simple, reliable metrics.[6,7]

Regardless of the ongoing debate around how to define and measure recovery,[3] we argue that there is a strong case to continue to explore clinical and functional outcomes of schizophrenia from an epidemiological perspective. In recent years, systematic reviews of the incidence,[8] prevalence,[9] and mortality of schizophrenia[10] have been published. Of the 4 key epidemiologic indicators required to understand the dynamics of disorders such as schizophrenia in a population (incidence, prevalence, remission/recovery, and mortality), recovery remains the most poorly understood. Clearly, the proportion of individuals who recover over a given period is more than 0% and (sadly) appears to be substantially less than 100%. Can we identify a range of values that encompass the best available estimates of recovery?

Several scholarly narrative reviews of outcome of schizophrenia have been published over recent decades.[11–14] While the definitions of remission and recovery have been the subject of a systematic review,[3] to the best of our knowledge, only 3 studies have examined the empirical data on "good outcomes" in schizophrenia using systematic reviews and/or meta-analytic techniques.[15–17]

According to a meta-analysis by Hegarty et al,[15] based of 320 studies published between 1895 and 1992, approximately 40% of schizophrenia patients were considered as having a good outcome. However, this review did not apply a minimum duration for good outcome, and it was acknowledged that the included studies used widely different methods to allocate subjects to the good outcome category. So, subjects showing either an improvement of symptoms or good social functioning may have been rated as having good outcome. Hegarty et al[15] found that studies using broad non-Kraepelinian diagnostic criteria had higher recovery percentages compared with Kraepelinian criteria, but the length of follow-up did not affect the proportion of recovery. Worryingly, this review noted that the proportion of patients with good outcome had not improved in recent decades. More recently, a systematic review by Menezes et al[16] of the outcome studies of first-episode psychosis was published. This review (based on 37 studies) concluded that 42% of patients had a good outcome. However, good outcome in this review did not require both good clinical and social/functional outcomes, and there was no requirement for good outcome status to have persisted for a certain period of time. This review was exclusively based on first-episode cases, with prospective follow-up for at least 6 months, though most samples had been followed-up for only a relatively short period (mean duration of follow-up was 35.1 months). Despite these caveats, the 2 systematic reviews reported remarkably similar proportions with good outcome (42% and 40%).

Warner[17] analyzed 114 follow-up studies (published between 1904 and 2000) to examine recovery in schizophrenia. He defined recovery as complete recovery (loss of psychotic symptoms and return to pre-illness level of functioning) or social recovery (economic and residential independence and low social disruption). No criterion for persistence of recovery was used. According to this analysis, 11%–33% were completely recovered and 22%–53% were socially recovered. Consistent with the findings of Hegarty et al,[15] with respect to changes in outcomes over time, Warner[17] reported that recovery rates had not increased over time.

Menezes and colleagues[16] recommended that multidimensional definitions should be used in future studies for outcome in psychosis. Other commentators have suggested that such multidimensional measures should include at least 2 domains—one related to clinical remission and another related to broader social functioning outcome.[18,19] Additionally, persistence of good outcome (indicating recovery) for a minimum of 2 years has been suggested.[18] Setting a duration criterion for persistent recovery does not exclude the possibility of relapses or continued recovery beyond that time. Mindful that not all outcome studies provide data on 2 or more domains and even fewer would share the same rating scales, we sought to collate the primary literature that reported outcome estimates based on these 2 domains. Previous reviews and original studies have often focused on cross-sectional outcomes, without any duration criteria. In this study, we will use the term recovery to describe very good outcome that considers both clinical and social/functional dimensions and includes a duration criteria of at least 2 years for at least 1 of these measures.

The broad objective of this study was to undertake a systematic review and meta-analysis of original studies reporting proportions of individuals with schizophrenia and related psychoses who met our predefined recovery criteria. Our primary aims were to identify the proportions of individuals who met our recovery criteria and to examine the nature of the distribution of these estimates (eg, median, mean estimate, and range).

We also examined potential sources of heterogeneity in the estimates in order to address selected research questions. For example, while narrative reviews have generally suggested that woman have better outcomes compared with men,[20] this issue was not addressed in the 3 previous systematic reviews. In recent years, there has been debate about the links between better clinical outcomes in schizophrenia and studies from sites with lower economic indices (often dichotomized to as "developing" or "developed" nations).[21,22] For example, Menezes et al[16] found better prognosis for samples from "developing country of origin." They also found an association between the methodological features of the study design and outcomes (eg, better outcome in studies with poorer representativeness and thus in studies with poorer quality). In light of the systematic reviews that have reported an association between reduced duration of untreated psychosis and better outcomes,[16,23,24] one might also predict that outcomes should have improved over time. In recent decades there has been increased focus on the detection and prompt treatment of early psychosis, which might result in better clinical outcomes.[25–27] Thus, optimistic researchers might predict that recovery proportions should improve over time as we continually attempt to improve treatments and service delivery. However, 2 earlier systematic reviews[15,17] found that the proportion of good outcomes had not improved over time—indeed there was evidence that estimates of good outcomes had declined in more recent studies. The lack of change in good outcomes over time is a finding that requires careful ongoing surveillance by the research community. We had the opportunity to reexamine this issue in this systematic review. Finally, we also wished to explore if a number of other design issues have an impact on recovery proportions (eg, first-episode status, narrow Kraepelinian definitions of schizophrenia, duration of follow-up, and study-quality score).

Mindful that the primary studies included in the review may not have been designed to examine these particular issues, our a priori hypotheses were the following:

  1. A greater proportion of women with schizophrenia and related psychoses would meet recovery criteria compared with men.

  2. The proportion of cases who recover will not have changed over time.

  3. A greater proportion of cases from studies from sites with poorer economic indices would meet recovery criteria compared with sites with richer economic indices.

  4. Recovery is more prevalent in first-episode samples compared with general samples.

  5. Recovery is more prevalent in samples using non-Kraepelinian vs Kraepelinian diagnostic system.

  6. Recovery is more prevalent in samples with longer duration of follow-up compared with shorter follow-up.

  7. Recovery is more prevalent in studies with lower quality scores.