1. Introduction
Age of onset is considered central to understanding the course of schizophrenia [Reference DeLisi1]. An earlier age of onset is associated with a worse course of the disorder and with poorer outcomes, including a worse course of psychiatric hospitalization in the long and short term [Reference Rabinowitz, Levine and Häfner2], poorer educational and vocational adjustment [Reference Lay, Blanz, Hartmann and Schmidt3], lower levels of social functioning and social cognition [Reference Linke, Jankowski, Ciołkiewicz, Jędrasik-Styła, Parnowska and Gruszka4], and an increased risk of suicide [Reference Kuo, Tsai, Lo, Wang and Chen5].
In contrast with most studies which reported that an earlier age of onset is associated with worse outcomes [Reference Rabinowitz, Levine and Häfner2–Reference Kuo, Tsai, Lo, Wang and Chen5], many studies report that age of onset is unrelated to quality of life [Reference Heinrichs, Hanlon and Carpenter6–Reference Bellino, Rocca, Patria, Marchiaro, Rasetti and Di Lorenzo10], a relevant outcome in schizophrenia [Reference Heinrichs, Hanlon and Carpenter6, Reference Strauss and Breier11]. Quality of life has developed into a central concept in mental health care [Reference Mashiach-Eizenberg, Hasson-Ohayon, Yanos, Lysaker and Roe12–Reference Boyer, Baumstarck, Boucekine, Blanc, Lançon and Auquier15] as treatment goals shifted from symptom change to incorporate recovery outcomes evaluating a personally meaningful life [Reference Liberman, Kopelowicz, Ventura and Gutkind16, Reference Priebe17]. Quality of life and symptom severity only moderately correlated [Reference Eack and Newhill18–Reference Phalen, Viswanadhan, Lysaker and Warman22] and so represent alternative approaches to outcome assessment. One approach views recovery as a subjective process while a contrasting approach defines recovery as an objective outcome [Reference Leonhardt, Huling, Hamm, Roe and Hasson-Ohayon23]. These contrasting approaches may account for the aforementioned particular results of a null association between age of onset and quality of life. Nonetheless, one study reported that early onset mediated associations with quality of life [Reference Roe, Mashiach-Eizenberg and Lysaker24], indicating that age of onset played a secondary role in quality of life. Thus a further examination of the role of age of onset on quality of life is needed.
The association between age of onset and quality of life is yet to be explored among males and females. Being female has been previously linked to lower quality of life in schizophrenia [Reference Huppert, Weiss, Lim, Pratt and Smith25], while being male has been linked to an earlier onset of schizophrenia [Reference Aleman, Kahn and Selten26, Reference Eranti, MacCabe, Bundy and Murray27] and poorer outcomes [Reference Aleman, Kahn and Selten26, Reference Angermeyer, Khn and Goldstein28, Reference Leung and Psych29]. These aforementioned findings may be accounted for by the distinctive nature of the association between age of onset and quality of life or by specific trends of this association, differentiating males and females.
Research suggests that mental health patients and professional caregivers usually appraise quality of life by relying on different aspects [Reference Bengtsson-Tops, Hansson, Sandlund, Bjarnason and Korkeila30–Reference Kusel, Laugharne, Perrington, McKendrick, Stephenson and Stockton-Henderson32]. Appraisals by professional caregivers are mostly based on psychiatric symptoms, whereas patients base their appraisals on physical health and social relations [Reference Bengtsson-Tops, Hansson, Sandlund, Bjarnason and Korkeila30, Reference Kravetz, Faust and Dasberg31]. Disparities between different rating sources are highly prevalent in studies of quality of life [Reference Voruganti, Heslegrave, Awad and Seeman13, Reference Bengtsson-Tops, Hansson, Sandlund, Bjarnason and Korkeila30–,Reference Kravetz, Faust and Dasberg31, Reference Kusel, Laugharne, Perrington, McKendrick, Stephenson and Stockton-Henderson32, Reference Doyle, Flanagan, Browne, Clarke, Lydon and Larkin33, Reference Hayhurst, Massie, Dunn, Lewis and Drake34, Reference Hasson-Ohayon, Roe, Kravetz, Levy-Frank and Meir35, Reference Ofir‐Eyal, Hasson‐Ohayon, Bar‐Kalifa and Kravetz36]. Hence both self-appraised and caregiver-appraised quality of life scores should be considered for robustness.
The current study aims to examine the association between the age of onset and quality of life, among males and females with schizophrenia, using a national study design with competing statistical models. The current study hypotheses are that age of onset would positively correlate with self-appraised and caregiver-appraised quality of life, and that this correlation would be amplified among males.
2. Methods
2.1. Participants
In this ongoing study, the current cohort was part of the Psychiatric Rehabilitation Routine Outcome Measurement Project [Reference Roe, Gelkopf, Gornemann, Baloush-Kleinman and Shadmi37]. The study was approved by the Helsinki committee at the Ministry of Health and by the Institutional Review Board at the University of Haifa. Assessments were conducted during psychiatric rehabilitation and were supervised by an internal service staff member or externally trained individuals [Reference Gelkopf, Pagorek-Eshel, Trauer and Roe38].
The current study included a subset of persons with a last diagnosis of schizophrenia who had information on birth year, dates of psychiatric hospital admissions and self and caregiver-appraised quality of life assessments (N = 1624). All participants received national psychiatric rehabilitation services in Israel. The exclusion criteria for psychiatric rehabilitation services are illicit drug addiction, violence and lack of psychiatric monitoring. Participants completed the research questionnaires from January 1st, 2013 to August 19th, 2015. Professional caregivers (N = 578) were given instruments that mirrored the one designed for self-appraisals.
2.2. Data sources
Demographic data (e.g. birth year), hospitalization information (e.g. date and duration of first and subsequent psychiatric hospital admissions) and psychiatric diagnosis were obtained from the Israeli National Psychiatric Case Registry. The registry, established in 1950, contains a lifelong listing of psychiatric hospitalizations in Israel, and includes ICD-10 diagnoses made by an Israeli medical board certified psychiatrist. Registry diagnoses include almost all persons with schizophrenia [Reference Weiser, Werbeloff, Dohrenwend, Levav, Yoffe and Davidson39], were found to be reliable over time [Reference Rabinowitz, Slyuzberg, Ritsner, Mark, Popper and Ginath40], and have acceptable sensitivity compared to research diagnostic criteria [Reference Weiser, Kanyas, Malaspina, Harvey, Glick and Goetz41]. The registry has been used in numerous studies [Reference Rabinowitz, Levine and Häfner2, Reference Levine, Lurie, Kohn and Levav42, Reference Ullman, Levine, Reichenberg and Rabinowitz43].
2.3. Quality of life appraisal
Mental health related quality of life was measured and validated [Reference Roe, Gelkopf, Gornemann, Baloush-Kleinman and Shadmi37] based on the Manchester Short Assessment of Quality of Life (MSA-QoL), an abbreviated version of the Lancaster Questionnaire Life Quality Profile [Reference Priebe, Huxley, Knight and Evans44]. Scale items were rated on a 5-point Likert scale and coded so higher scores on the assessment indicated better quality of life. Eight items measured satisfaction with one’s work or volunteering activities, financial status, social status and activities, family relations, leisure activities, residential status, physical health condition and mental health condition (one item per life area). Quality of life appraisals for persons with schizophrenia were made by two rating sources: self-appraised and by their professional caregivers. Caregivers were given an instrument that mirrored the one designed for self-appraisals. The internal reliabilities of the measure were acceptable for self and caregiver-appraisals (α = 0.76; α = 0.77, respectively).
2.4. Analytic approach
First, descriptive analyses were computed of the sample characteristics. A modest yet statistically significant correlation was observed between self-appraised and caregiver-appraised quality of life (r = 0.29, p < 0.01), like prior research [Reference Bengtsson-Tops, Hansson, Sandlund, Bjarnason and Korkeila30–Reference Ofir‐Eyal, Hasson‐Ohayon, Bar‐Kalifa and Kravetz36]. Mean scores were significantly (t(1623) = 24.01, p <.05) higher for self-appraisals (M = 28.86, SD = 5.59) compared to caregiver appraisals (M = 25.12, SD = 4.89), like prior research [Reference Ofir‐Eyal, Hasson‐Ohayon, Bar‐Kalifa and Kravetz36]. Hence, self-appraised and caregiver-appraised quality of life scores were analyzed for robustness.
Second, the primary statistical analysis examined the association between self-appraised and then caregiver-appraised quality of life as a function of age of onset using regression models. The assumptions of the regression models were tested. Visual inspection of residual figures was performed in order to reveal deviations from homoscedasticity or normality. An inspection for normality of error terms followed using a histogram and probability plots of the residuals. Independence of the error term was examined through a scatter plot of residuals by the predicted values to show that no discernible association existed. Then, statistically significant outliers were removed as their inclusion offsets estimation [Reference Tabachnick and Fidell45]. Next, multiple regression analysis models were computed for self-appraised quality of life. Mixed models with random intercepts for caregivers were then used to account for the repeated caregiver ratings in the data. Significance for mixed regression models was calculated by comparing models via the likelihood ratio test.
Competing models for each rating source were computed in ascending complexity and tested without adjustment and adjusted for confounding of birth year, age at the time of data collection, duration of disorder and whether male or female. The age of onset was derived from the difference between the date of birth and date of first psychiatric hospital admission. Duration of disorder was derived from the difference between the time of data collection and the date of first psychiatric hospital admission and was categorized into four categories cased on quartile scores. Age at the time of data collection was derived from the difference between the time of data collection and the date of birth and was categorized into two categories based on the median score. Birth year was categorized into two categories based on the median score. Models for each rating source were numbered as follows. The first model accounted for a linear effect of age of onset on self-appraised quality of life (model 1 hereafter). The second model accounted for a linear effect of age of onset, whether male or female, birth year, age at the time of data collection and duration of disorder (model 2 hereafter). The third model accounted for a quadratic effect of age of onset on self-appraised quality of life (model 3 hereafter). The fourth model accounted for a quadratic effect of age of onset, whether male or female, birth year, age at the time of data collection and duration of disorder (model 4 hereafter). The fifth model accounted for a cubic effect of age of onset on self-appraised quality of life (model 5 hereafter). The sixth model accounted for a cubic effect of age of onset, whether male or female, birth year, age at the time of data collection and duration of disorder (model 6 hereafter).
Third, models for each rating source were compared for parsimony based on the Bayesian Information Criterion (BIC) for model selection [Reference Schwarz46], similar to prior research [Reference Furukawa, Levine, Tanaka, Goldberg, Samara and Davis47, Reference Samara, Goldberg, Levine, Furukawa, Geddes and Cipriani48]. Lower BIC values represent more parsimonious models and so are a better fit to the data.
In instances of quadratic models, the cutoff point was computed by setting the second derivative equal to zero [Reference Hazewinkel49]. Since an upward concave corresponds to a positive second derivative and downward concave corresponds to a negative second derivative, then when the function changes from upward concave to downward concave (or vice-versa) the second derivative equals to zero.
Sensitivity analyses were conducted for males and females, like in prior research [Reference Rabinowitz, Levine and Häfner2]. All regression models were recomputed separately for males and females and then compared for parsimony.
All analyses were computed in R [Reference R Core Team50] with the lme4 library [Reference Bates, Maechler and Bolker51] used to compute linear mixed effects of the associations between caregiver-appraised quality of life and age of onset.
3. Results
3.1. Sample characteristics
The final sample was based on 1551 persons with schizophrenia rated by 564 professional caregivers. This sample size was a result of removing all outliers (N = 73, 4.5%). Of the remainder a total of 62% (N = 957) of the participants were males (N = 594 females). Overall mean age for first psychiatric hospital admission was 25.55 years (SD = 8.32), 26.13 years (SD = 8.91) for females, and 25.18 years (SD = 7.92) for males. Overall mean birth year was 1967 (SD = 12.03 years; mean age at time of data collection was 46.5 years), for females 1966 (SD = 12.13 years; mean age at time of data collection was 47.84 years), and for males 1968 (SD = 11.90 years; mean age at time of data collection was 45.66 years). Overall mean duration of disorder was 21.05 years (SD = 11.72), 21.81 years (SD = 12.02) for females, and 20.58 years (SD = 11.52) for males. See Table 1 for mean and 95% confidence interval values of quality of life appraisals and age of onset.
Among the caregivers, 44% (N = 249) appraised one person with schizophrenia, 21% (N = 120) appraised two persons, 11% (N = 61) appraised three persons, and 24% (N = 134) appraised four or more persons.
3.2. Primary statistical analysis
3.2.1. Self-appraised quality of life as a function of age of onset
Comparison of the six competing regression models showed that model 1 was the most parsimonious (BIC = 9573.99; Table 2; See Table 3 for comparison of BIC values for all self-appraised quality of life models). This model indicated that age of onset alone was significantly (P <.01) negatively associated with self-appraised quality of life (Fig. 1a).
Table 1 Mean and 95% confidence interval values of quality of life and age of onset.
Table 2 Best-fitting quality of life model parameters.
Self-appraised quality of life models:
Model 1: Self-appraised quality of life = Intercept + Age of Onset.
Model 1- F: Female self-appraised quality of life = Intercept + Female Age of Onset.
Model 3- M: Male self-appraised quality of life = Intercept + Male Age of Onset + (Male Age of Onset)2.
Caregiver-appraised quality of life models:
Model 1: Caregiver-appraised quality of life = Intercept + Age of Onset.
Model 1- F: Caregiver-appraised female quality of life = Intercept + Female Age of Onset.
3.2.2. Caregiver-appraised quality of life as a function of age of onset
Comparison of the six competing linear mixed effect models showed that model 1 was the most parsimonious (BIC = 9291.33; Table 2; See Table 3 for comparison of BIC values for all caregiver-appraised quality of life models). This model accounted for a linear effect of age of onset on caregiver-appraised quality of life, beyond the effect of each individual caregiver (Fig. 2a) and indicated that age of onset alone was significantly (P <.05) negatively associated with caregiver-appraised quality of life.
Table 3 BIC values for self-appraised and caregiver-appraised quality of life models.
Note: The lowest BIC is the best-fitting model and is underlined for clarity.
Self-appraised quality of life models:
Model 1: Self-appraised Quality of Life = Intercept + Age of Onset.
Model 2: Self-appraised Quality of Life = Intercept + Age of Onset + Birth year + Duration of disorder + Age.
Model 3: Self-appraised Quality of Life = Intercept + Age of Onset + (Age of Onset)2.
Model 4: Self-appraised Quality of Life = Intercept + Age of Onset + (Age of Onset)2 + Birth year + Duration of disorder + Age.
Model 5: Self-appraised Quality of Life = Intercept + Age of Onset + (Age of Onset)2 + (Age of Onset)3.
Model 6: Self-appraised Quality of Life = Intercept + Age of Onset + (Age of Onset)2 + (Age of Onset)3+ Birth year + Duration of disorder + Age.
Model 2, model 4 and model 6 for the total sample were also adjusted for being male or female.
Caregiver-appraised quality of life models:
Model 1: Caregiver-appraised Quality of Life = Intercept + Age of Onset.
Model 2: Caregiver -appraised Quality of Life = Intercept + Age of Onset + Birth year + Duration of disorder + Age.
Model 3: Caregiver -appraised Quality of Life = Intercept + Age of Onset + (Age of Onset)2.
Model 4: Caregiver -appraised Quality of Life = Intercept + Age of Onset + (Age of Onset)2+ Birth year + Duration of disorder + Age.
Model 5: Caregiver -appraised Quality of Life = Intercept + Age of Onset + (Age of Onset)2 + (Age of Onset)3.
Model 6: Caregiver -appraised Quality of Life = Intercept + Age of Onset + (Age of Onset)2 + (Age of Onset)3+ Birth year + Duration of disorder + Age.
Model 2, model 4 and model 6 for the total sample were also adjusted for being male or female.
3.3. Sensitivity analysis: age of onset and quality of life among males and females
All aforementioned models were recomputed separately for males and females (see Table 3 for BIC values). Consistent with the results of the primary analysis, among females with schizophrenia, for self-appraised quality of life, model 1 (a linear effect of age of onset on self-appraised quality of life) was the best fitting model (BIC = 3750.57; Table 2; Fig. 1b). Among males with schizophrenia, model 3 was the best fitting model (BIC = 6012.84; Table 2; Fig. 1c). Model 3 consisted of a quadratic effect of age of onset alone on self-appraised quality of life, inconsistent with the linear effect found for the total sample. This model had an inflection point, in which the curve changed from being concave downward to being concave upward, at onset age of 36.67 years.
The unique effects of males and females were also found for the mixed models (see Table 3 for BIC values). Consistent with the primary analysis, among females with schizophrenia, for caregiver-appraised quality of life, model 1 (a linear effect of age of onset beyond the effect of individual caregivers, on caregivers’ appraisals of quality of life) was the best fitting model (BIC = 3607.00; Table 2; Fig. 2b). Among males with schizophrenia, age of onset could not significantly predict caregiver-appraised quality of life, using any of the mixed models described (see Fig. 2c), a result which was inconsistent with the linear effect found for the total sample.
4. Discussion
The current study examined the association between the age of onset and quality of life, among males and females with schizophrenia, using data from a national sample and competing statistical models. Results show that self-appraised or caregiver-appraised quality of life decreased as age of onset increased, contrary to the study hypothesis of a positive correlation.
Fig. 1. Self-appraised quality of life as a function of age of onset.
Note: The slope in figure 1a is based on Model 1 (Self-appraised quality of life = Intercept + Age of Onset).
The slope in figure 1b is based on Model 1 for females (Self-appraised female quality of life = Intercept + Female Age of Onset).
The slope in figure 1c is based on Model 3 for males (Self-appraised male quality of life = Intercept + Male Age of Onset + (Male Age of Onset)2.
Fig. 2. Caregiver-appraised quality of life as a function of age of onset.
Note: The slope in figure 2a is based on Model 1 (Caregiver-appraised quality of life = Intercept + Age of Onset).
The slope in figure 2b is based on Model 1 for females (Caregiver-appraised female quality of life = Intercept Female Age of Onset).
Figure 2c presents a non-significant slope since for males, age of onset was not statistically significantly associated with caregiver-appraised quality of life using any of the mixed models computed.
The current study result that age of onset is negatively associated with quality of life in schizophrenia, may seem incongruent with much research claiming that earlier onset age is associated with poorer outcomes [Reference Rabinowitz, Levine and Häfner2]. Nonetheless, quality of life differs in its experience [Reference Heinrichs, Hanlon and Carpenter6] compared to the other outcomes (e.g., the course of psychiatric hospitalizations) that have been examined for their association with age of onset. Taken together with prior research, social decline may tentatively explain the reported tendency to generally appraise lower quality of life as the age of onset increases [Reference Harris, Cullum and Jeste52–Reference Häfner, Maurer, Löffler, Hambrecht and Schultze-Lutter56].
Similarly, the current findings show that for females with schizophrenia, quality of life appraisals decreased as the age of onset increased. Some earlier evidence suggests that a unique trend for females may be accounted for by psychiatric symptom severity. Late-onset females with schizophrenia (but not males) were found to have a general negative symptoms score more unfavorable than early-onset females [Reference Häfner, Hambrecht, Löffler, Munk-Jørgensen and Riecher-Rössler53, Reference Häfner, Maurer, Löffler and Riecher-R57], perhaps due to a reduced protective effect of estrogen [Reference Häfner54]. Psychiatric symptoms were steadily found to be negatively related to quality of life (see meta-analysis (18)) and their severity may tentatively explain the progressing decrease in female quality of life.
Finally, among males, results showed a quadratic effect of onset age on self-appraised quality of life, contrary to the study hypothesis that the positive correlation between onset age and quality of life would be amplified. The quadratic effect demonstrated a negative association up to onset age of 36.67 years, after which the association was positive. The age of 36 has been observed to distinguish early and late onset [Reference Vinokur, Levine, Roe, Krivoy and Fischel58] and confer to a reduced risk of suicide, especially for males [Reference Levine, Bakst and Rabinowitz59]. The positive association between age of onset and quality of life among males after the age of 36.67 may reflect the higher chances of retaining acquired social advantages associated with later onset [Reference Häfner, Hambrecht, Löffler, Munk-Jørgensen and Riecher-Rössler53]. Male onset age was not associated with caregiver-appraised quality of life, suggesting that the association is not robust like previously discussed, and that disparities between different rating sources largely exist for males.
5. Limitations
The current results are based on an abridged version of the MANSA [Reference Priebe, Huxley, Knight and Evans44]. More recent and broader quality of life measures exist that may yield different trends. Further, the underlying concept of quality of life in this questionnaire is generic and not disorder-specific. The measure, however, was chosen for pragmatic reasons of time and administration to a large sample [Reference Rotstein, Roe, Gelkopf, Shadmi and Levine60]. Additionally, although past studies suggest that information from patients is no less valid than data from other sources [Reference Strauss, Carpenter and Nasrallah61], the current data consist of two quality of life rating sources which enable greater confidence than had a single data source been used. Finally, quality of life was appraised by different sources than those from which onset age was derived, and there was a difference of 21.05 years (on average, SD = 11.72) between the actual age of onset and the time of assessment completion. Still, future research is warranted to ensure that the results generalize to other tools and samples.
Other than date of first psychiatric hospital admission, different age of onset measures, such as age when behavioral changes begin to occur or age at first manifestation of positive symptoms (e.g.(1)), were not available. However, past research indicates that there is close correspondence between different onset definitions [Reference DeLisi, Goldin, Maxwell, Kazuba and Gershon62, Reference DeLisi, Hoff, Schwartz, Shields, Halthore and Gupta63].
Furthermore, a longer duration of untreated psychosis, which was not measured in the current study, cannot be ruled out as an explanation of the association found in this study. Indeed, other research has reported a moderate yet significant correlation between duration of untreated psychosis and onset age [Reference Ho, Andreasen, Flaum, Nopoulos and Miller64, Reference Hastrup, Haahr, Jansen and Simonsen65].
Similarly, the current study did not consist of information on the psychiatric symptoms of schizophrenia disorder (e.g., the PANSS [Reference Kay, Fiszbein and Opfer66]) and specifically on negative symptoms which were found integral to quality of life [Reference Häfner, Hambrecht, Löffler, Munk-Jørgensen and Riecher-Rössler53, Reference Häfner, Maurer, Löffler and Riecher-R57]. We were therefore unable to learn of the actual interplay between symptoms, age of onset and quality of life, among males and females. Future studies may look into these associations.
6. Summary
This is the first study to examine competing statistical models of the association between quality of life appraisals and age of onset using a national study design. Age of onset was significantly associated with quality of life in schizophrenia, whether self-appraised or caregiver-appraised, and this association differed among males and females. Quality of life appraisals decreased as the age of onset increased for the general sample and for females. For males, a quadratic association was found between age of onset and self-appraised quality of life. Results may be tentatively explained by mechanisms of social decline, social advantages and symptomatology.
Funding
This publication is based on the results of the Outcomes Project that was conducted by Professor David Roe and Professor Marc Gelkopf of The Center for Mental Health Research, Practice and Policy at the Department of Community Mental Health with joint funding from the Israeli Ministry of Health and the Laszlo N. Tauber Family Foundation.
Conflict of interest
None. The authors report no conflict of interest in the past three years and are responsible for the content and writing of the paper.
Contributors
Author Anat Rotstein wrote the publication study protocol, managed the literature searches, conducted the statistical analysis and wrote the first draft of the manuscript. Authors Anat Rotstein and SZL designed the publication study protocol. All authors provided critical statistical and manuscript feedback. All authors contributed to and approved the final manuscript.
Acknowledgements
We are grateful acknowledge for the ongoing collaboration with the Mental Health Rehabilitation Department, especially Ronit Dudai and Vered Balush, and the Department of Information and Evaluation at the Ministry of Health, especially Rinat Yoffe as well as the entire team of the Psychiatric Rehabilitation Patient Reported Outcome Measurement project, notably Aid Rohanna, Paula Garber, Adi Telem and Ronit Mirkin. We acknowledge with thanks the support of The Center for Mental Health Research, Practice and Policy at the University of Haifa.
Comments
No Comments have been published for this article.