A Systematic Review of Recovery in Schizophrenia
A Systematic Review of Recovery in Schizophrenia
The electronic database searches identified 5647 unique records. After further screening, we identified 917 potential records. Figure 1 shows the PRISMA flow diagram that details the filtering process of potential studies. From these, 37 articles or books met all our criteria and were included in the systematic review. These 37 articles or books included altogether 50 discrete samples, including 13 samples from the WHO incidence and prevalence cohorts (including also unpublished data from Dr Kim Hopper) and 7 samples found from manual search. In total, these studies included 8994 discrete individuals and were derived from 20 different countries.
(Enlarge Image)
Figure 1.
Flow diagram of the selection of studies of recovery in schizophrenia.
Supplementary Material Table 1 summarizes key features of the included studies' study design, sample characteristics, the location where the study was conducted, diagnostic system, length of the follow-up, criteria for recovery, and recovery estimates. The citations for these articles are included in the online supplementary material Appendix 4.
For persons (males and females combined), we identified 50 estimates. The distribution of these estimates is shown in figure 2. Based on this distribution, the median recovery estimate was 13.5% (mean: 16.4%) with the IQR between 8.1% and 20.0%. The distribution was densely underpinned with estimates in its central 75% portion and was left-skewed (ie, some studies reported very high estimates). The mean estimates changed only slightly in the "influence analyses," when 1 study was dropped at a time (for these analyses, the estimates ranged between 15.5% and 16.7%). As expected, we confirmed that estimates from the included studies were highly heterogeneous (I = 99.8%; Q = 38 000, P < .001). The median annual recovery rate was 1.4% per annum (IQR: 0.7%–2.6%). With this annual recovery percentage, over 10 years approximately 14% would be expected to recover.
(Enlarge Image)
Figure 2.
Recovery percentage for included studies.
For sex-specific estimates (12 studies for males; 12 studies for females), the median recovery estimate for males was 12.9% (IQR: 10.0%–19.4%), while for females the median recovery estimate was 12.1% (IQR: 7.5%–29.0%; Table 1 ). Only 10 studies reported recovery percentage by gender, and for these studies, the OR for gender was calculated (figure 3). Eight out of 11 studies reported slightly higher recovery estimate for males, while 3 studies showed higher recovery for females, and 2 of these studies showed clearly better recovery for females; overall, there was no statistically significant difference between sexes (OR: 1.02, 95% CI: 0.62,1.69, P = .939).
(Enlarge Image)
Figure 3.
Odds ratios for gender differences in original studies.
When ranked according to the year at the midpoint of the data collection period using the same year categories as was employed by Warner, the median recovery was 13.0% in studies with midpoint before 1941, 17.7% in studies between 1941 and 1955, 16.9% in 1956–1975, 9.9% in 1976–1995, and 6.0% in studies after 1996 (P = .704; Table 1 ).
Compared with countries with high and upper middle income, recovery estimate was significantly higher in low or lower middle-income countries (medians 13.0% in high income countries, 12.1% in upper-middle, and 36.4% in low or lower middle-income countries) (t = 2,93, P = .005; Table 1 ). When this analysis was adjusted by the middle point of the collecting of the study, the difference remained statistically significant (t = −3.86, P < .001). When WHO studies were excluded from the crude analyses, the median recovery percentages were 15.0% (n = 34), 9.7% (n = 1), and 12.7% (n = 2), respectively, (t = 0.48, P = .632).
The recovery estimate was numerically but not signi- ficantly lower in studies using Kraepelinian diagnostic system (median 9.0%) compared with non-Kraepelinian samples (12.5%) (P = .396; Table 1 ). There were no significant differences in recovery when studies were classified according to first-episode studies vs general intake (P = .857), origin of the sample (P = .802), duration of follow-up (P = .369), being a WHO study vs not (P = .185), and quality score (P = .792; Table 1 ).
When the estimates were ranked according to the strictness of the recovery criteria, the recovery percentage in studies in the lower half, ie, less strict criteria (n = 25) was 13.0% (IQR: 8.1%–23.0%), while studies with stricter criteria (n = 25) had a median recovery proportion of 13.9% (IQR: 8.7–19.0%) (t = −1.15, P = .254). The score for strictness of the definition of recovery is presented for each study in online Supplementary Material Table 1. In 39 studies where only the clinical recovery criteria has lasted for at least 2 years, the median recovery was 13.0% (8.1–25.2%), whereas in 3 studies where only the social/functional recovery has lasted for at least 2 years, recovery was 19.0% (16.3–19.7%). In 8 studies where both clinical and social/functional recovery have lasted at least 2 years, recovery was 13.1% (7.7–16.8%). The chances of recovery are quite similar in studies where the required duration for clinical or functional recovery is under 5 years (median recovery 13.2%, IQR: 8.4%–24.1%; n = 28) vs over 5 years (14.7%, 7.3%–18.4%; n = 22).
Results
The electronic database searches identified 5647 unique records. After further screening, we identified 917 potential records. Figure 1 shows the PRISMA flow diagram that details the filtering process of potential studies. From these, 37 articles or books met all our criteria and were included in the systematic review. These 37 articles or books included altogether 50 discrete samples, including 13 samples from the WHO incidence and prevalence cohorts (including also unpublished data from Dr Kim Hopper) and 7 samples found from manual search. In total, these studies included 8994 discrete individuals and were derived from 20 different countries.
(Enlarge Image)
Figure 1.
Flow diagram of the selection of studies of recovery in schizophrenia.
Supplementary Material Table 1 summarizes key features of the included studies' study design, sample characteristics, the location where the study was conducted, diagnostic system, length of the follow-up, criteria for recovery, and recovery estimates. The citations for these articles are included in the online supplementary material Appendix 4.
For persons (males and females combined), we identified 50 estimates. The distribution of these estimates is shown in figure 2. Based on this distribution, the median recovery estimate was 13.5% (mean: 16.4%) with the IQR between 8.1% and 20.0%. The distribution was densely underpinned with estimates in its central 75% portion and was left-skewed (ie, some studies reported very high estimates). The mean estimates changed only slightly in the "influence analyses," when 1 study was dropped at a time (for these analyses, the estimates ranged between 15.5% and 16.7%). As expected, we confirmed that estimates from the included studies were highly heterogeneous (I = 99.8%; Q = 38 000, P < .001). The median annual recovery rate was 1.4% per annum (IQR: 0.7%–2.6%). With this annual recovery percentage, over 10 years approximately 14% would be expected to recover.
(Enlarge Image)
Figure 2.
Recovery percentage for included studies.
For sex-specific estimates (12 studies for males; 12 studies for females), the median recovery estimate for males was 12.9% (IQR: 10.0%–19.4%), while for females the median recovery estimate was 12.1% (IQR: 7.5%–29.0%; Table 1 ). Only 10 studies reported recovery percentage by gender, and for these studies, the OR for gender was calculated (figure 3). Eight out of 11 studies reported slightly higher recovery estimate for males, while 3 studies showed higher recovery for females, and 2 of these studies showed clearly better recovery for females; overall, there was no statistically significant difference between sexes (OR: 1.02, 95% CI: 0.62,1.69, P = .939).
(Enlarge Image)
Figure 3.
Odds ratios for gender differences in original studies.
When ranked according to the year at the midpoint of the data collection period using the same year categories as was employed by Warner, the median recovery was 13.0% in studies with midpoint before 1941, 17.7% in studies between 1941 and 1955, 16.9% in 1956–1975, 9.9% in 1976–1995, and 6.0% in studies after 1996 (P = .704; Table 1 ).
Compared with countries with high and upper middle income, recovery estimate was significantly higher in low or lower middle-income countries (medians 13.0% in high income countries, 12.1% in upper-middle, and 36.4% in low or lower middle-income countries) (t = 2,93, P = .005; Table 1 ). When this analysis was adjusted by the middle point of the collecting of the study, the difference remained statistically significant (t = −3.86, P < .001). When WHO studies were excluded from the crude analyses, the median recovery percentages were 15.0% (n = 34), 9.7% (n = 1), and 12.7% (n = 2), respectively, (t = 0.48, P = .632).
The recovery estimate was numerically but not signi- ficantly lower in studies using Kraepelinian diagnostic system (median 9.0%) compared with non-Kraepelinian samples (12.5%) (P = .396; Table 1 ). There were no significant differences in recovery when studies were classified according to first-episode studies vs general intake (P = .857), origin of the sample (P = .802), duration of follow-up (P = .369), being a WHO study vs not (P = .185), and quality score (P = .792; Table 1 ).
When the estimates were ranked according to the strictness of the recovery criteria, the recovery percentage in studies in the lower half, ie, less strict criteria (n = 25) was 13.0% (IQR: 8.1%–23.0%), while studies with stricter criteria (n = 25) had a median recovery proportion of 13.9% (IQR: 8.7–19.0%) (t = −1.15, P = .254). The score for strictness of the definition of recovery is presented for each study in online Supplementary Material Table 1. In 39 studies where only the clinical recovery criteria has lasted for at least 2 years, the median recovery was 13.0% (8.1–25.2%), whereas in 3 studies where only the social/functional recovery has lasted for at least 2 years, recovery was 19.0% (16.3–19.7%). In 8 studies where both clinical and social/functional recovery have lasted at least 2 years, recovery was 13.1% (7.7–16.8%). The chances of recovery are quite similar in studies where the required duration for clinical or functional recovery is under 5 years (median recovery 13.2%, IQR: 8.4%–24.1%; n = 28) vs over 5 years (14.7%, 7.3%–18.4%; n = 22).