1. |
|
|
2. |
- Chen, Qi, et al.
(author)
-
Predicting suicide attempt or suicide death following a visit to psychiatric specialty care : A machine learning study using Swedish national registry data
- 2020
-
In: PLoS Medicine. - San Francisco : Public Library of Science (PLoS). - 1549-1277 .- 1549-1676. ; 17:11
-
Journal article (peer-reviewed)abstract
- BACKGROUND: Suicide is a major public health concern globally. Accurately predicting suicidal behavior remains challenging. This study aimed to use machine learning approaches to examine the potential of the Swedish national registry data for prediction of suicidal behavior.METHODS AND FINDINGS: The study sample consisted of 541,300 inpatient and outpatient visits by 126,205 Sweden-born patients (54% female and 46% male) aged 18 to 39 (mean age at the visit: 27.3) years to psychiatric specialty care in Sweden between January 1, 2011 and December 31, 2012. The most common psychiatric diagnoses at the visit were anxiety disorders (20.0%), major depressive disorder (16.9%), and substance use disorders (13.6%). A total of 425 candidate predictors covering demographic characteristics, socioeconomic status (SES), electronic medical records, criminality, as well as family history of disease and crime were extracted from the Swedish registry data. The sample was randomly split into an 80% training set containing 433,024 visits and a 20% test set containing 108,276 visits. Models were trained separately for suicide attempt/death within 90 and 30 days following a visit using multiple machine learning algorithms. Model discrimination and calibration were both evaluated. Among all eligible visits, 3.5% (18,682) were followed by a suicide attempt/death within 90 days and 1.7% (9,099) within 30 days. The final models were based on ensemble learning that combined predictions from elastic net penalized logistic regression, random forest, gradient boosting, and a neural network. The area under the receiver operating characteristic (ROC) curves (AUCs) on the test set were 0.88 (95% confidence interval [CI] = 0.87-0.89) and 0.89 (95% CI = 0.88-0.90) for the outcome within 90 days and 30 days, respectively, both being significantly better than chance (i.e., AUC = 0.50) (p < 0.01). Sensitivity, specificity, and predictive values were reported at different risk thresholds. A limitation of our study is that our models have not yet been externally validated, and thus, the generalizability of the models to other populations remains unknown.CONCLUSIONS: By combining the ensemble method of multiple machine learning algorithms and high-quality data solely from the Swedish registers, we developed prognostic models to predict short-term suicide attempt/death with good discrimination and calibration. Whether novel predictors can improve predictive performance requires further investigation.
|
|
3. |
- Dobrosavljevic, Maja, 1986-, et al.
(author)
-
Risk prediction model for cardiovascular diseases in adults initiating pharmacological treatment for attention-deficit/hyperactivity disorder
- 2022
-
In: Evidence-Based Mental Health. - : BMJ Publishing Group Ltd. - 1362-0347 .- 1468-960X. ; 25, s. 185-190
-
Journal article (peer-reviewed)abstract
- Background: Available prediction models ofcardiovascular diseases (CVDs) may not accuratelypredict outcomes among individuals initiatingpharmacological treatment for attention-deficit/hyperactivity disorder (ADHD).Objective: To improve the predictive accuracyof traditional CVD risk factors for adults initiatingpharmacological treatment of ADHD, by consideringnovel CVD risk factors associated with ADHD (comorbidpsychiatric disorders, sociodemographic factors andpsychotropic medication).Methods: The cohort composed of 24 186 adultsresiding in Sweden without previous CVDs, born between1932 and 1990, who started pharmacological treatmentof ADHD between 2008 and 2011, and were followedfor up to 2 years. CVDs were identified using diagnosesaccording to the International Classification of Diseases,and dispended medication prescriptions from Swedishnational registers. Cox proportional hazards regressionwas employed to derive the prediction model.Findings: The developed model included eighttraditional and four novel CVD risk factors. Themodel showed acceptable overall discrimination (Cindex=0.72, 95% CI 0.70 to 0.74) and calibration(Brier score=0.008). The Integrated DiscriminationImprovement index showed a significant improvementafter adding novel risk factors (0.003 (95% CI 0.001 to0.007), p<0.001).Conclusions: The inclusion of the novel CVD riskfactors may provide a better prediction of CVDs in thispopulation compared with traditional CVD predictorsonly, when the model is used with a continuous riskscore. External validation studies and studies assessingclinical impact of the model are warranted.Clinical implications: Individuals initiatingpharmacological treatment of ADHD at higher risk ofdeveloping CVDs should be more closely monitored.
|
|
4. |
- Fazel, Seena, et al.
(author)
-
Risk of death by suicide following self-harm presentations to healthcare : development and validation of a multivariable clinical prediction rule (OxSATS)
- 2023
-
In: BMJ Mental Health. - : BMJ Publishing Group Ltd. - 2755-9734. ; 26:1
-
Journal article (peer-reviewed)abstract
- BACKGROUND: Assessment of suicide risk in individuals who have self-harmed is common in emergency departments, but is often based on tools developed for other purposes. OBJECTIVE: We developed and validated a predictive model for suicide following self-harm.METHODS: We used data from Swedish population-based registers. A cohort of 53 172 individuals aged 10+ years, with healthcare episodes of self-harm, was split into development (37 523 individuals, of whom 391 died from suicide within 12 months) and validation (15 649 individuals, 178 suicides within 12 months) samples. We fitted a multivariable accelerated failure time model for the association between risk factors and time to suicide. The final model contains 11 factors: age, sex, and variables related to substance misuse, mental health and treatment, and history of self-harm. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis guidelines were followed for the design and reporting of this work.FINDINGS: An 11-item risk model to predict suicide was developed using sociodemographic and clinical risk factors, and showed good discrimination (c-index 0.77, 95% CI 0.75 to 0.78) and calibration in external validation. For risk of suicide within 12 months, using a 1% cut-off, sensitivity was 82% (75% to 87%) and specificity was 54% (53% to 55%). A web-based risk calculator is available (Oxford Suicide Assessment Tool for Self-harm or OxSATS).CONCLUSIONS: OxSATS accurately predicts 12-month risk of suicide. Further validations and linkage to effective interventions are required to examine clinical utility.CLINICAL IMPLICATIONS: Using a clinical prediction score may assist clinical decision-making and resource allocation.
|
|
5. |
- Ghirardi, Laura, et al.
(author)
-
Neurodevelopmental disorders and subsequent risk of violent victimization : exploring sex differences and mechanisms
- 2022
-
In: Psychological Medicine. - : Cambridge University Press. - 0033-2917 .- 1469-8978. ; 53:4, s. 1510-1517
-
Journal article (peer-reviewed)abstract
- Background: Neurodevelopmental disorders (NDs) are associated with experiences of victimization, but mechanisms remain unclear. We explored sex differences and the role of familial factors and externalizing problems in the association between several NDs and violent victimization in adolescence and young adulthood.Methods: Individuals born in Sweden 1985-1997, residing in Sweden at their 15th birthday, were followed until date of violent victimization causing a hospital visit or death, death due to other causes, emigration, or December 31, 2013, whichever came first. The exposures were diagnoses of attention-deficit/hyperactivity disorder (ADHD), autism spectrum disorder (ASD), intellectual disability (ID) and other NDs. We used three different Cox regression models: a crude model, a model adjusted for familial confounding using sibling-comparisons, and a model additionally adjusted for externalizing problems.Results: Among 1 344 944 individuals followed, on average, for 5 years, 74 487 were diagnosed with NDs and 37 765 had a hospital visit or died due to violence. ADHD was associated with an increased risk of violent victimization in males [hazard ratio (HR) 2.56; 95% confidence interval (CI) 2.43-2.70) and females (HR 5.39; 95% CI 4.97-5.85). ASD and ID were associated with an increased risk of violent victimization in females only. After adjusting for familial factors and externalizing problems, only ADHD was associated with violent victimization among males (HR 1.27; 95% CI 1.06-1.51) and females (HR 1.69; 95% CI 1.21-2.36).Conclusions: Females with NDs and males with ADHD are at greater risk of being victim of severe violence during adolescence and young adulthood. Relevant mechanisms include shared familial liability and externalizing problems. ADHD may be independently associated with violent victimization.
|
|
6. |
- Hegvik, Tor-Arne, et al.
(author)
-
Labor epidural analgesia and subsequent risk of offspring autism spectrum disorder and attention-deficit/hyperactivity disorder : A cross-national cohort study of 4.5 million individuals and their siblings
- 2023
-
In: American Journal of Obstetrics and Gynecology. - : Elsevier. - 0002-9378 .- 1097-6868. ; 228:2, s. 233.e1-233.e12
-
Journal article (peer-reviewed)abstract
- BACKGROUND: A recent study has suggested that labor epidural analgesia may be associated with increased rates of offspring autism spectrum disorder (ASD). Subsequent replication attempts have lacked sufficient power to confidently exclude the possibility of a small effect and the causal nature of this association remains unknown.OBJECTIVE: To investigate the extent to which exposure to labor epidural analgesia is associated with offspring ASD and attention-deficit/hyperactivity disorder (ADHD) following adjustments for unmeasured familial confounding.STUDY DESIGN: We identified 4,498,462 singletons and their parents using the Medical Birth Registers in Finland (cohorts born 1987-2005), Norway (1999-2015), and Sweden (1987-2011), linked with population and patient registries. These cohorts were followed from birth until they either had the outcomes of interest, emigrated, died, or reached the end of the follow-up (at mean ages 13.6-16.8 years), whichever occurred first. Cox regression models were used to estimate country-specific associations between labor epidural analgesia recorded at birth and outcomes (e.g., at least one secondary care diagnosis of ASD and ADHD or at least one dispensed prescription of medication used for the treatment of ADHD). The models were adjusted for sex, birth year, birth order, and unmeasured familial confounders via sibling-comparisons. Pooled estimates across all three countries were estimated using inverse variance weighted fixed-effects meta-analysis models.RESULTS: A total of 4,498,462 individuals (48.7% female) were included, 1,091,846 (24.3%) of which were exposed to labor epidural analgesia. Of these, 1.2% were diagnosed with ASD and 4.0% with ADHD. On the population level, pooled estimates showed that labor epidural analgesia was associated with increased risk of offspring ASD (adjusted hazard ratio, aHR=1.12; 95% CI: 1.10-1.14, absolute risks: 1.20% vs. 1.07%) and ADHD (aHR=1.20; 1.19-1.21; 3.95% vs. 3.32%). However, when comparing full-siblings who were differentially exposed to labor epidural analgesia, the associations were fully attenuated for both conditions, with narrow confidence intervals (aHRASD=0.98; 0.93-1.03; aHRADHD=0.99; 0.96-1.02).CONCLUSION: In this large cross-national study, we found no support for the hypothesis that exposure to labor epidural analgesia causes either offspring autism spectrum disorder or attention-deficit/hyperactivity disorder.
|
|
7. |
- Latvala, Antti, et al.
(author)
-
Association of intellectual disability with violent and sexual crime and victimization : a population-based cohort study
- 2023
-
In: Psychological Medicine. - Cambridge : Cambridge University Press. - 0033-2917 .- 1469-8978. ; 35:9, s. 1-9
-
Journal article (peer-reviewed)abstract
- BACKGROUND: Intellectual disability (ID) is associated with violent and sexual offending and victimization, but the importance of neuropsychiatric comorbidity and severity of disability remains unclear.METHODS: In a register-based cohort study of people born in Sweden 1980-1991 (n = 1 232 564), we investigated associations of mild and moderate/severe ID with any, violent and sexual crimes, and with assault victimization, stratified by comorbid autism and attention deficit hyperactivity disorder (ADHD). We defined ID by attendance at a special school or registered diagnosis and obtained data on criminal convictions and injuries or deaths due to assaults from nationwide registers until end of 2013.RESULTS: Compared to people without ID, autism or ADHD, men and women with mild or moderate/severe ID and comorbid ADHD had elevated risks of violent crimes [range of hazard ratios (HRs) 4.4-10.4] and assault victimization (HRs 2.0-7.7). Women with mild ID without comorbidities or with comorbid autism also had elevated risks of violent crimes and victimization (HRs 1.8-4.6) compared to women without ID, autism or ADHD. The relative risks of sexual offending and victimization were elevated in men and women with ID without comorbidities (HRs 2.6-12.7). The highest risks for sexual offending in men (HRs 9.4-11.0) and for sexual assault victimization in women (HRs 11.0-17.1) related to ID and comorbid ADHD.CONCLUSIONS: The elevated risk of violent offending and assault victimization in people with ID is largely explained by comorbid ADHD, whereas ID is independently associated with sexual crimes and victimization, even though absolute risks are low.
|
|
8. |
- Molero, Yasmina, et al.
(author)
-
Associations between statin use and suicidality, depression, anxiety, and seizures : a Swedish total-population cohort study
- 2020
-
In: Lancet psychiatry. - : Elsevier. - 2215-0374 .- 2215-0366. ; 7:11, s. 982-990
-
Journal article (peer-reviewed)abstract
- BACKGROUND: Statins have shown both protective and adverse associations with neuropsychiatric outcomes. We aimed to examine the possible associations between statins and suicidality, depression, anxiety, and seizures.METHODS: Using Swedish national registers, we linked data on dispensed statin prescriptions with data on unplanned (emergency) hospital visits or specialised outpatient care for four neuropsychiatric outcomes: suicidal behaviour (including deaths from suicide), depressive disorders, anxiety disorders, and seizures. We included all individuals in the registries who were dispensed statins and who were aged 15 years or older between Jan 1, 2006, and Dec 31, 2013. We applied a within-individual design using stratified Cox proportional hazards regression to compare the incidence of the defined outcomes during periods on statins and periods off statins within each individual, thus adjusting for time-invariant confounders. Non-specific effects of treatment were tested by investigating these outcomes in relation to thiazide diuretic use and antihistamine use in the same cohort.FINDINGS: The statin-users cohort comprised 1 149 384 individuals, of whom 1 015 949 (88·4%) were aged 50 years or older, 625 616 (54·4%) were male, and 523 768 (45·6%) were female. The study period consisted of 2 053 310 non-treatment periods and 2 997 545 treatment periods, and 957 216 (83·3%) individuals had a medication status change (from on statins to off statins, or vice versa). Suicide outcomes were found in 6372 (0·6%) individuals, depressive disorders in 23 745 (2·1%), anxiety disorders in 30 100 (2·6%), and seizures in 28 844 (2·5%). There were no clear associations between periods of statin treatment and suicidal behaviour or deaths from suicide (hazard ratio 0·99 [95% CI 0·90-1·08]), anxiety disorders (0·99 [0·95-1·02]), or seizures (1·00 [0·97-1·04]). Statins were associated with reduced hazards of depressive disorders (0·91 [0·87-0·94]), which remained after adjustment for concurrent antidepressant use (0·91 [0·88-0·94]). Hazard ratios for depressive disorders were 0·61 (0·38-1·00; n=14 718) with thiazide diuretic use and 0·84 (0·67-1·06; n=23 715) with antihistamine use.INTERPRETATION: Statin use is not associated with suicidality, anxiety disorders, or seizures. Whether the observed association between statin use and reduced diagnoses of clinical depression is confounded by non-specific benefits related to being prescribed medication needs further research.
|
|
9. |
- Molero, Yasmina, et al.
(author)
-
Associations between β-blockers and psychiatric and behavioural outcomes : A population-based cohort study of 1.4 million individuals in Sweden
- 2023
-
In: PLoS Medicine. - : Public Library of Science (PLoS). - 1549-1277 .- 1549-1676. ; 20:1
-
Journal article (peer-reviewed)abstract
- BACKGROUND: β-blockers are widely used for treating cardiac conditions and are suggested for the treatment of anxiety and aggression, although research is conflicting and limited by methodological problems. In addition, β-blockers have been associated with precipitating other psychiatric disorders and suicidal behaviour, but findings are mixed. We aimed to examine associations between β-blockers and psychiatric and behavioural outcomes in a large population-based cohort in Sweden.METHODS AND FINDINGS: We conducted a population-based longitudinal cohort study using Swedish nationwide high-quality healthcare, mortality, and crime registers. We included 1,400,766 individuals aged 15 years or older who had collected β-blocker prescriptions and followed them for 8 years between 2006 and 2013. We linked register data on dispensed β-blocker prescriptions with main outcomes, hospitalisations for psychiatric disorders (not including self-injurious behaviour or suicide attempts), suicidal behaviour (including deaths from suicide), and charges of violent crime. We applied within-individual Cox proportional hazards regression to compare periods on treatment with periods off treatment within each individual in order to reduce possible confounding by indication, as this model inherently adjusts for all stable confounders (e.g., genetics and health history). We also adjusted for age as a time-varying covariate. In further analyses, we adjusted by stated indications, prevalent users, cardiac severity, psychiatric and crime history, individual β-blockers, β-blocker selectivity and solubility, and use of other medications. In the cohort, 86.8% (n = 1,215,247) were 50 years and over, and 52.2% (n = 731,322) were women. During the study period, 6.9% (n = 96,801) of the β-blocker users were hospitalised for a psychiatric disorder, 0.7% (n = 9,960) presented with suicidal behaviour, and 0.7% (n = 9,405) were charged with a violent crime. There was heterogeneity in the direction of results; within-individual analyses showed that periods of β-blocker treatment were associated with reduced hazards of psychiatric hospitalisations (hazard ratio [HR]: 0.92, 95% confidence interval [CI]: 0.91 to 0.93, p < 0.001), charges of violent crime (HR: 0.87, 95% CI: 0.81 to 0.93, p < 0.001), and increased hazards of suicidal behaviour (HR: 1.08, 95% CI: 1.02 to 1.15, p = 0.012). After stratifying by diagnosis, reduced associations with psychiatric hospitalisations during β-blocker treatment were mainly driven by lower hospitalisation rates due to depressive (HR: 0.92, 95% CI: 0.89 to 0.96, p < 0.001) and psychotic disorders (HR: 0.89, 95% CI: 0.85 to 0.93, p < 0.001). Reduced associations with violent charges remained in most sensitivity analyses, while associations with psychiatric hospitalisations and suicidal behaviour were inconsistent. Limitations include that the within-individual model does not account for confounders that could change during treatment, unless measured and adjusted for in the model.CONCLUSIONS: In this population-wide study, we found no consistent links between β-blockers and psychiatric outcomes. However, β-blockers were associated with reductions in violence, which remained in sensitivity analyses. The use of β-blockers to manage aggression and violence could be investigated further.
|
|
10. |
- Molero, Yasmina, et al.
(author)
-
Medication utilization in traumatic brain injury patients-insights from a population-based matched cohort study
- 2024
-
In: Frontiers in Neurology. - : Frontiers Media S.A.. - 1664-2295. ; 15
-
Journal article (peer-reviewed)abstract
- INTRODUCTION: Traumatic brain injury (TBI) is associated with health problems across multiple domains and TBI patients are reported to have high rates of medication use. However, prior evidence is thin due to methodological limitations. Our aim was thus to examine the use of a wide spectrum of medications prescribed to address pain and somatic conditions in a population-based cohort of TBI patients, and to compare this to a sex- and age-matched cohort. We also examined how patient factors such as sex, age, and TBI severity were associated with medication use.METHODS: We assessed Swedish nationwide registers to include all individuals treated for TBI in hospitals or specialist outpatient care between 2006 and 2012. We examined dispensed prescriptions for eight different non-psychotropic medication classes for the 12 months before, and 12 months after, the TBI. We applied a fixed-effects model to compare TBI patients with the matched population cohort. We also stratified TBI patients by sex, age, TBI severity and carried out comparisons using a generalized linear model.RESULTS: We identified 239,425 individuals with an incident TBI and 239,425 matched individuals. TBI patients were more likely to use any medication [Odds ratio (OR) = 2.03, 95% Confidence Interval (CI) = 2.00-2.05], to present with polypharmacy (OR = 1.96, 95% CI = 1.90-2.02), and to use each of the eight medication classes before their TBI, as compared to the matched population cohort. Following the TBI, TBI patients were more likely to use any medication (OR = 1.83, 95% CI = 1.80-1.86), to present with polypharmacy (OR = 1.74, 95% CI = 1.67-1.80), and to use all medication classes, although differences were attenuated. However, differences increased for antibiotics/antivirals (OR = 2.02, 95% CI = 1.99-2.05) and NSAIDs/antirheumatics (OR = 1.62, 95% CI = 1.59-1.65) post-TBI. We also found that females and older patients were more likely to use medications after their TBI than males and younger patients, respectively. Patients with more severe TBIs demonstrated increased use of antibiotics/ antivirals and NSAIDs/antirheumatics than those with less severe TBIs.DISCUSSION: Taken together, our results point to poor overall health in TBI patients, suggesting that medical follow-up should be routine, particularly in females with TBI, and include a review of medication use to address potential polypharmacy.
|
|