External validation of two prediction models identifying employees at risk of high sickness absence: cohort study with 1-year follow-up

Predicting long-term sickness absence with employee questionnaires and administrative records: a prospective cohort study of hospital employees

OBJECTIVE

This study aimed to compare the utility of risk estimation derived from questionnaires and administrative records in predicting long-term sickness absence among shift workers.

METHODS

This prospective cohort study comprised 3197 shift-working hospital employees (mean age 44.5 years, 88.0% women) who responded to a brief 8-item questionnaire on work disability risk factors and were linked to 28 variables on their working hour and workplace characteristics obtained from administrative registries at study baseline. The primary outcome was the first sickness absence lasting ≥90 days during a 4-year follow-up.

RESULTS

The C-index of 0.73 [95% confidence interval (CI) 0.70-0.77] for a questionnaire-only based prediction model, 0.71 (95% CI 0.67-0.75) for an administrative records-only model, and 0.79 (95% CI 0.76-0.82) for a model combining variables from both data sources indicated good discriminatory ability. For a 5%-estimated risk as a threshold for positive test results, the detection rates were 76%, 74%, and 75% and the false positive rates were 40%, 45% and 34% for the three models. For a 20%-risk threshold, the corresponding detection rates were 14%, 8%, and 27% and the false positive rates were 2%, 2%, and 4%. To detect one true positive case with these models, the number of false positive cases accompanied varied between 7 and 10 using the 5%-estimated risk, and between 2 and 3 using the 20%-estimated risk cut-off. The pattern of results was similar using 30-day sickness absence as the outcome.

CONCLUSIONS

The best predictive performance was reached with a model including both questionnaire responses and administrative records. Prediction was almost as accurate with models using only variables from one of these data sources. Further research is needed to examine the generalizability of these findings.

Development and internal validation of a nomogram to predict massive blood transfusions in neurosurgical operations

Objectives

A massive blood transfusion (MBT) is an unexpected event that may impact mortality. Neurosurgical operations are a major operation involving the vital structures and risk to bleeding. The aims of the present research were (1) to develop a nomogram to predict MBT and (2) to estimate the association between MBT and mortality in neurosurgical operations.

Material and Method

We conducted a retrospective cohort study including 3660 patients who had undergone neurosurgical operations. Univariate and multivariate logistic regression analyses were used to test the association between clinical factors, pre-operative hematological laboratories, and MBT. A nomogram was developed based on the independent predictors.

Results

The predictive model comprised five predictors as follows: Age group, traumatic brain injury, craniectomy operation, pre-operative hematocrit, and pre-operative international normalized ratio and the good calibration were observed in the predictive model. The concordance statistic index was 0.703. Therefore, the optimism-corrected c-index values of cross-validation and bootstrapping were 0.703 and 0.703, respectively.

Conclusion

MBT is an unexpectedly fatal event that should be considered for appropriate preparation blood components. Further, this nomogram can be implemented for allocation in limited-resource situations in the future.

Prediction of bowel management independence after ischemic spinal cord injury

BACKGROUND

Ischemic spinal cord injury (SCI) belongs to the heterogeneous group of non-traumatic SCI, while the course of sensorimotor and functional recovery is comparable to traumatic SCI. Recently, we derived from data of patients with traumatic SCI a valid model to predict an independent and reliable bowel management one year after SCI.

AIM

To evaluate the performance of this model to predict an independent and reliable bowel management one year following ischemic SCI.

DESIGN

Prognostic study - observational study.

SETTING

European Multicenter Study about Spinal Cord Injury (EMSCI) ClinicalTrials.gov: NCT01571531.

POPULATION

One hundred and forty-two patients with ischemic SCI of various level and severity of injury.

METHODS

The prediction model relied on a single predictor collected within 40 days from injury, the International Standards for Neurological Classification of Spinal Cord Injury total motor score. Bowel outcome one year after SCI derived from the dichotomization of the Spinal Cord Independence Measure (SCIM) item 7 scores. We defined a positive outcome as independent bowel management with regular movements and appropriate timing with no or rare accidents (score of 10 in SCIM version II and score of 8 or 10 in version III).

RESULTS

The model showed a fair discrimination with an area under the receiver operating characteristic (ROC) curve of 0.780 (95% confidence interval=0.702-0.860). In addition, the model displayed an acceptable accuracy and calibration.

CONCLUSIONS

The study extends the validity of our rule to patients with ischemic SCI, thus providing the first model to predict an independent and reliable bowel management in this population.

CLINICAL REHABILITATION IMPACT

The model may be employed in clinical practice to counsel patients, to define the rehabilitation aims and to estimate the need of assistance after discharge, as well as in the research field for the optimization of patients' allocation in the design of future clinical trials.

Using machine learning for the personalised prediction of revision endoscopic sinus surgery

BACKGROUND

Revision endoscopic sinus surgery (ESS) is often considered for chronic rhinosinusitis (CRS) if maximal conservative treatment and baseline ESS prove insufficient. Emerging research outlines the risk factors of revision ESS. However, accurately predicting revision ESS at the individual level remains uncertain. This study aims to examine the prediction accuracy of revision ESS and to identify the effects of risk factors at the individual level.

METHODS

We collected demographic and clinical variables from the electronic health records of 767 surgical CRS patients ≥16 years of age. Revision ESS was performed on 111 (14.5%) patients. The prediction accuracy of revision ESS was examined by training and validating different machine learning models, while the effects of variables were analysed using the Shapley values and partial dependence plots.

RESULTS

The logistic regression, gradient boosting and random forest classifiers performed similarly in predicting revision ESS. Area under the receiving operating characteristic curve (AUROC) values were 0.744, 0.741 and 0.730, respectively, using data collected from the baseline visit until six months after baseline ESS. The length of time during which data were collected improved the prediction performance. For data collection times of 0, 3, 6 and 12 months after baseline ESS, AUROC values for the logistic regression were 0.682, 0.715, 0.744 and 0.784, respectively. The number of visits before or after baseline ESS, the number of days from the baseline visit to the baseline ESS, patient age, CRS with nasal polyps (CRSwNP), asthma, non-steroidal anti-inflammatory drug exacerbated respiratory disease and immunodeficiency or suspicion of it all associated with revision ESS. Patient age and number of visits before baseline ESS carried non-linear effects for predictions.

CONCLUSIONS

Intelligent data analysis identified important predictors of revision ESS at the individual level, such as the frequency of clinical visits, patient age, Type 2 high diseases and immunodeficiency or a suspicion of it.

Development of a nomogram to predict the outcome of moderate or severe pediatric traumatic brain injury

OBJECTIVES:

Traumatic brain injury (TBI) in children has become the major cause of mortality and morbidity in Thailand that has had an impact with economic consequences. This study aimed to develop and internally validate a nomogram for a 6-month follow-up outcome prediction in moderate or severe pediatric TBI.

METHODS:

This retrospective cohort study involved 104 children with moderate or severe TBI. Various clinical variables were reviewed. The functional outcome was assessed at the hospital discharge and at a 6-month follow-up based on the King's Outcome Scale for Childhood Head Injury classification. Predictors associated with the 6-month follow-up outcome were developed from the predictive model using multivariable binary logistic regression to estimate the performance and internal validation. A nomogram was developed and presented as a predictive model.

RESULTS:

The mean age of the samples was 99.75 months (standard deviation 59.65). Road traffic accidents were the highest injury mechanism at 84.6%. The predictive model comprised Glasgow Coma Scale of 3–8 (odds ratio [OR]: 16.07; 95% confidence interval [CI]: 1.27–202.42), pupillary response in one eye (OR 7.74; 95% CI 1.26–47.29), pupillary nonresponse in both eyes (OR: 57.74; 95% CI: 2.28–145.81), hypotension (OR: 19.54; 95%: CI 3.23–117.96), and subarachnoid hemorrhage (OR: 9.01, 95% CI: 1.33–60.80). The concordance statistic index (C-index) of the model's discrimination was 0.931, while the C-index following the bootstrapping and 5-cross validation were 0.920 and 0.924, respectively.

CONCLUSIONS:

The performance of a clinical nomogram for predicting 6-month follow-up outcomes in pediatric TBI patients was assessed at an excellent level. However, further external validation would be required for the confirmation of the tool's performance.

Application of machine learning to predict the outcome of pediatric traumatic brain injury

PURPOSE

Traumatic brain injury (TBI) generally causes mortality and disability, particularly in children. Machine learning (ML) is a computer algorithm, applied as a clinical prediction tool. The present study aims to assess the predictability of ML for the functional outcomes of pediatric TBI.

METHODS

A retrospective cohort study was performed targeting children with TBI who were admitted to the trauma center of southern Thailand between January 2009 and July 2020. The patient was excluded if he/she (1) did not undergo a CT scan of the brain, (2) died within the first 24 h, (3) had unavailable complete medical records during admission, or (4) was unable to provide updated outcomes. Clinical and radiologic characteristics were collected such as vital signs, Glasgow coma scale score, and characteristics of intracranial injuries. The functional outcome was assessed using the King's Outcome Scale for Childhood Head Injury, which was thus dichotomized into favourable outcomes and unfavourable outcomes: good recovery and moderate disability were categorized as the former, whereas death, vegetative state, and severe disability were categorized as the latter. The prognostic factors were estimated using traditional binary logistic regression. By data splitting, 70% of data were used for training the ML models and the remaining 30% were used for testing the ML models. The supervised algorithms including support vector machines, neural networks, random forest, logistic regression, naive Bayes and k-nearest neighbor were performed for training of the ML models. Therefore, the ML models were tested for the predictive performances by the testing datasets.

RESULTS

There were 828 patients in the cohort. The median age was 72 months (interquartile range 104.7 months, range 2-179 months). Road traffic accident was the most common mechanism of injury, accounting for 68.7%. At hospital discharge, favourable outcomes were achieved in 97.0% of patients, while the mortality rate was 2.2%. Glasgow coma scale score, hypotension, pupillary light reflex, and subarachnoid haemorrhage were associated with TBI outcomes following traditional binary logistic regression; hence, the 4 prognostic factors were used for building ML models and testing performance. The support vector machine model had the best performance for predicting pediatric TBI outcomes: sensitivity 0.95, specificity 0.60, positive predicted value 0.99, negative predictive value 1.0; accuracy 0.94, and area under the receiver operating characteristic curve 0.78.

CONCLUSION

The ML algorithms of the present study have a high sensitivity; therefore they have the potential to be screening tools for predicting functional outcomes and counselling prognosis in general practice of pediatric TBIs.

External Validation of the Bone Metastases Ensemble Trees for Survival (BMETS) Machine Learning Model to Predict Survival in Patients With Symptomatic Bone Metastases

PURPOSE

The Bone Metastases Ensemble Trees for Survival (BMETS) model uses a machine learning algorithm to estimate survival time following consultation for palliative radiation therapy for symptomatic bone metastases (SBM). BMETS was developed at a tertiary-care, academic medical center, but its validity and stability when applied to external data sets are unknown.

PATIENTS AND METHODS

Patients treated with palliative radiation therapy for SBM from May 2013 to May 2016 at two hospital-based community radiation oncology clinics were included, and medical records were retrospectively reviewed to collect model covariates and survival time. The Kaplan-Meier method was used to estimate overall survival from consultation to death or last follow-up. Model discrimination was estimated using time-dependent area under the curve (tAUC), which was calculated using survival predictions from BMETS based on the initial training data set.

RESULTS

A total of 216 sites of SBM were treated in 182 patients. Most common histologies were breast (27%), lung (23%), and prostate (23%). Compared with the BMETS training set, the external validation population was older (mean age, 67 v 62 years; P < .001), had more primary breast (27% v 19%; P = .03) and prostate cancer (20% v 12%; P = .01), and survived longer (median, 10.7 v 6.4 months). When the BMETS model was applied to the external data set, tAUC values at 3, 6, and 12 months were 0.82, 0.77, and 0.77, respectively. When refit with data from the combined training and external validation sets, tAUC remained > 0.79.

CONCLUSION

BMETS maintained high discriminative ability when applied to an external validation set and when refit with new data, supporting its generalizability, stability, and the feasibility of dynamic modeling.

Predicting Long-Term Sickness Absence and Identifying Subgroups Among Individuals Without an Employment Contract

Purpose Today, decreasing numbers of workers in Europe are employed in standard employment relationships. Temporary contracts and job insecurity have become more common. This study among workers without an employment contract aimed to (i) predict risk of long-term sickness absence and (ii) identify distinct subgroups of sick-listed workers. Methods 437 individuals without an employment contract who were granted a sickness absence benefit for at least two weeks were followed for 1 year. We used registration data and self-reported questionnaires on sociodemographics, work-related, health-related and psychosocial factors. Both were retrieved from the databases of the Dutch Social Security Institute and measured at the time of entry into the benefit. We used logistic regression analysis to identify individuals at risk of long-term sickness absence. Latent class analysis was used to identify homogenous subgroups of individuals. Results Almost one-third of the study population (n = 133; 30%) was still at sickness absence at 1-year follow-up. The final prediction model showed fair discrimination between individuals with and without long-term sickness absence (optimism adjusted AUC to correct for overfitting = 0.761). Four subgroups of individuals were identified based on predicted risk of long-term sickness absence, self-reported expectations about recovery and return to work, reason of sickness absence and coping skills. Conclusion The logistic regression model could be used to identify individuals at risk of long-term sickness absence. Identification of risk groups can aid professionals to offer tailored return to work interventions.

Long-term sickness absence in a working population: development and validation of a risk prediction model in a large Dutch prospective cohort

BACKGROUND

Societal expenditures on work-disability benefits is high in most Western countries. As a precursor of long-term work restrictions, long-term sickness absence (LTSA) is under continuous attention of policy makers. Different healthcare professionals can play a role in identification of persons at risk of LTSA but are not well trained. A risk prediction model can support risk stratification to initiate preventative interventions. Unfortunately, current models lack generalizability or do not include a comprehensive set of potential predictors for LTSA. This study is set out to develop and validate a multivariable risk prediction model for LTSA in the coming year in a working population aged 45-64 years.

METHODS

Data from 11,221 working persons included in the prospective Study on Transitions in Employment, Ability and Motivation (STREAM) conducted in the Netherlands were used to develop a multivariable risk prediction model for LTSA lasting ≥28 accumulated working days in the coming year. Missing data were imputed using multiple imputation. A full statistical model including 27 pre-selected predictors was reduced to a practical model using backward stepwise elimination in a logistic regression analysis across all imputed datasets. Predictive performance of the final model was evaluated using the Area Under the Curve (AUC), calibration plots and the Hosmer-Lemeshow (H&L) test. External validation was performed in a second cohort of 5604 newly recruited working persons.

RESULTS

Eleven variables in the final model predicted LTSA: older age, female gender, lower level of education, poor self-rated physical health, low weekly physical activity, high self-rated physical job load, knowledge and skills not matching the job, high number of major life events in the previous year, poor self-rated work ability, high number of sickness absence days in the previous year and being self-employed. The model showed good discrimination (AUC 0.76 (interquartile range 0.75-0.76)) and good calibration in the external validation cohort (H&L test: p = 0.41).

CONCLUSIONS

This multivariable risk prediction model distinguishes well between older workers with high- and low-risk for LTSA in the coming year. Being easy to administer, it can support healthcare professionals in determining which persons should be targeted for tailored preventative interventions.

Developing and validating models for predicting nursing home admission using only RAI-HC instrument data

OBJECTIVE

In recent years research has identified important predictors for nursing home admission (NHA). However, as far as we know, the previous risk models use complex variable sets from many sources and the output is a single risk value. The objective of this study was to develop an NHA risk model with a variable set from single data source and richer output information.

METHODS

In this study, we developed a model selecting variables only from the RAI-HC (Resident Assessment Instrument - Home Care) system. Furthermore, we used principal component analysis and K-means clustering to target proper interventions for high-risk clients.

RESULTS

The performance of the model was close to the complex previous model (recall [Formula: see text] vs. [Formula: see text] and specificity [Formula: see text] vs. [Formula: see text]). For the risk clients, three intervention clusters (deficiency in physical functionality, deficiency in cognitive functionality and depression and mood disorders) were found.

CONCLUSION

The NHA risk model and intervention clusters are important because they enable the identification of proper interventions for the right clients. The fact that the model with RAI-HC data alone was accurate enough simplifies the integration of the NHA risk model into practice because it uses data from one system and the algorithm can be integrated easily into the source system.

Risk factors associated with oral manifestations and oral health impact of gastro-oesophageal reflux disease: a multicentre, cross-sectional study in Pakistan

OBJECTIVE

Gastro-oesophageal reflux disease (GORD) is a relatively common disorder and manifests with extraoesophageal symptoms, such as dental erosions (DE), cough, laryngitis, asthma, and oral soft- and hard-tissue pathologies. This study aimed (1) to identify oral soft and hard-tissue changes in patients with GORD and (2) to evaluate these oral changes as indices for assessing GORD and its severity.

SETTING

This cross-sectional study was conducted at four major tertiary care government hospitals, in two metropolitan cities of Pakistan.

PARTICIPANTS

In total, 187 of 700 patients who underwent oesophago-gastro-duodenoscopy and having GORD were included in the study. Patients with GORD were divided according to the presence of DE into group A (with DE, chronic/severe GORD) and group B (without DE, mild GORD). Patients who were unconscious and had extremely limited mouth opening were excluded.

PRIMARY AND SECONDARY OUTCOME MEASURES

Abnormal conditions and lesions of the oral mucosa were recorded. The impact of oral hard and soft-tissue changes on the oral health-related quality of life was assessed using the Pakistani (Urdu) version of the validated Oral Health Impact Profile-14 (OHIP-14) instrument.

RESULTS

Oral submucous fibrosis (66.3%), ulceration (59.4%) and xerostomia (47.6%) were significantly more common in group A (p<0.05). The prevalence of GORD was 26.7%, within which the prevalence of DE was 35.3%. Unhealthy dietary pattern, nausea/vomiting, oesophagitis, xerostomia, ulceration, gingivitis and angular cheilitis showed a statistically significant association with chronic GORD and DE. All subscales of OHIP-14 were positively correlated (p<0.05) in patients with GORD and DE, with notable impact on psychological discomfort (rs=0.30), physical disability (rs=0.29), psychological disability (rs=0.27) and functional limitation (rs=0.20).

CONCLUSION

Patients with GORD and DE presented with more severe oral manifestations than did those with GORD and no DE. We recommend timely dental check-ups to assess the severity of both systemic and oral disease.

Host Taxon Predictor - A Tool for Predicting Taxon of the Host of a Newly Discovered Virus

Recent advances in metagenomics provided a valuable alternative to culture-based approaches for better sampling viral diversity. However, some of newly identified viruses lack sequence similarity to any of previously sequenced ones, and cannot be easily assigned to their hosts. Here we present a bioinformatic approach to this problem. We developed classifiers capable of distinguishing eukaryotic viruses from the phages achieving almost 95% prediction accuracy. The classifiers are wrapped in Host Taxon Predictor (HTP) software written in Python which is freely available at https://github.com/wojciech-galan/viruses_classifier. HTP’s performance was later demonstrated on a collection of newly identified viral genomes and genome fragments. In summary, HTP is a culture- and alignment-free approach for distinction between phages and eukaryotic viruses. We have also shown that it is possible to further extend our method to go up the evolutionary tree and predict whether a virus can infect narrower taxa.

Development and validation of classifiers and variable subsets for predicting nursing home admission

Background

In previous years a substantial number of studies have identified statistically important predictors of nursing home admission (NHA). However, as far as we know, the analyses have been done at the population-level. No prior research has analysed the prediction accuracy of a NHA model for individuals.

Methods

This study is an analysis of 3056 longer-term home care customers in the city of Tampere, Finland. Data were collected from the records of social and health service usage and RAI-HC (Resident Assessment Instrument - Home Care) assessment system during January 2011 and September 2015. The aim was to find out the most efficient variable subsets to predict NHA for individuals and validate the accuracy. The variable subsets of predicting NHA were searched by sequential forward selection (SFS) method, a variable ranking metric and the classifiers of logistic regression (LR), support vector machine (SVM) and Gaussian naive Bayes (GNB). The validation of the results was guaranteed using randomly balanced data sets and cross-validation. The primary performance metrics for the classifiers were the prediction accuracy and AUC (average area under the curve).

Results

The LR and GNB classifiers achieved 78% accuracy for predicting NHA. The most important variables were RAI MAPLE (Method for Assigning Priority Levels), functional impairment (RAI IADL, Activities of Daily Living), cognitive impairment (RAI CPS, Cognitive Performance Scale), memory disorders (diagnoses G30-G32 and F00-F03) and the use of community-based health-service and prior hospital use (emergency visits and periods of care).

Conclusion

The accuracy of the classifier for individuals was high enough to convince the officials of the city of Tampere to integrate the predictive model based on the findings of this study as a part of home care information system. Further work need to be done to evaluate variables that are modifiable and responsive to interventions.

Comparison of six outcome prediction models in an adult burn population in a developing country

There are two types of prognostic model - burn-specific and general - to predict mortality risk in burn patients. Most prediction models were devised in developed countries. The aim of this study was to compare the performance of six outcome models in a developing country. In a retrospective cohort study, data of all thermal burned adult patients (age ≥ 18 years) admitted to the Burn Intensive Care Unit (BICU) were collected and then the following six prediction models were used to assess each patient: Acute Physiology and Chronic Health Evaluation (APACHE II), Abbreviated Burn Severity Index (ABSI), Belgian Outcome in Burn Injury (BOBI), the Ryan model, revised Baux and FLAMES model. Discriminative ability and goodness-of-fit of the prediction models were determined by receiver operating characteristic curve analysis and Hosmer-Lemeshow tests. We included 238 patients (mean age: 38.3 ± 18.39 years, average TBSA: 58.27% ± 24.55) in our study; 172 (72.3%) of them were diagnosed with inhalation injury and 178 (72.4%) were intubated. Mortality rate was 69.7%. Deceased patients had significantly higher mean age, %TBSA and number of inhalation injury. The area under the curve of the models was between 64.5 (APACHE II) and 85.9 (ABSI). The best estimation of predicted mortality was obtained with the ABSI model (67.2%).

Focus Group Study Exploring Factors Related to Frequent Sickness Absence

Introduction

Research investigating frequent sickness absence (3 or more episodes per year) is scarce and qualitative research from the perspective of frequent absentees themselves is lacking. The aim of the current study is to explore awareness, determinants of and solutions to frequent sickness absence from the perspective of frequent absentees themselves.

Methods

We performed a qualitative study of 3 focus group discussions involving a total of 15 frequent absentees. Focus group discussions were audiotaped and transcribed verbatim. Results were analyzed with the Graneheim method using the Job Demands Resources (JD–R) model as theoretical framework.

Results

Many participants were not aware of their frequent sickness absence and the risk of future long-term sickness absence. As determinants, participants mentioned job demands, job resources, home demands, poor health, chronic illness, unhealthy lifestyles, and diminished feeling of responsibility to attend work in cases of low job resources. Managing these factors and improving communication (skills) were regarded as solutions to reduce frequent sickness absence.

Conclusions

The JD–R model provided a framework for determinants of and solutions to frequent sickness absence. Additional determinants were poor health, chronic illness, unhealthy lifestyles, and diminished feeling of responsibility to attend work in cases of low job resources. Frequent sickness absence should be regarded as a signal that something is wrong. Managers, supervisors, and occupational health care providers should advise and support frequent absentees to accommodate job demands, increase both job and personal resources, and improve health rather than express disapproval of frequent sickness absence and apply pressure regarding work attendance.

Diagnostic accuracy of the Patient Health Questionnaire-9 for assessment of depression in type II diabetes mellitus and/or coronary heart disease in primary care

BACKGROUND

Depression is common among type 2 diabetes mellitus (DM2)/coronary heart disease (CHD) patients and is associated with adverse health effects. A promising strategy to reduce burden of disease is to identify patients at risk for depression in order to offer indicated prevention. This study aims to assess the diagnostic accuracy of the Patient Health Questionnaire-9 (PHQ-9) to be used as a tool to identify high risk patients.

METHODS

In this cross-sectional study, 586 consecutive DM2/CHD patients aged >18 were recruited through 23 general practices. PHQ-9 outcomes were compared to the Mini International Neuropsychiatric Interview (MINI), which was considered the reference standard. Diagnostic accuracy was evaluated for minor and major depression, comparing both sum- and algorithm based PHQ-9 scores.

RESULTS

For minor depression, the optimal cut-off score was 8 (sensitivity 71%, specificity 71% and an AUC of 0.74). For major depression, the optimal cut-off score was 10 resulting in a sensitivity of 84%, a specificity of 82%, and an AUC of 0.88. The positive predictive value of the PHQ-9 algorithm for diagnosing minor and major depression was 25% and 33%, respectively.

LIMITATIONS

Two main limitations apply. MINI Interviewers were not blinded for PHQ-9 scores and less than 10% of all invited patients could be included in the analyses. This could have resulted in biased outcomes.

CONCLUSIONS

The PHQ-9 sum score performs well in identifying patients at high risk of minor and major depression. However, the PHQ-9 showed suboptimal results for diagnostic purposes. Therefore, it is recommended to combine the use of the PHQ-9 with further diagnostics to identify depression.

Cross-national validation of prognostic models predicting sickness absence and the added value of work environment variables

PURPOSE

To validate Dutch prognostic models including age, self-rated health and prior sickness absence (SA) for ability to predict high SA in Danish eldercare. The added value of work environment variables to the models' risk discrimination was also investigated.

METHODS

2,562 municipal eldercare workers (95% women) participated in the Working in Eldercare Survey. Predictor variables were measured by questionnaire at baseline in 2005. Prognostic models were validated for predictions of high (≥30) SA days and high (≥3) SA episodes retrieved from employer records during 1-year follow-up. The accuracy of predictions was assessed by calibration graphs and the ability of the models to discriminate between high- and low-risk workers was investigated by ROC-analysis. The added value of work environment variables was measured with Integrated Discrimination Improvement (IDI).

RESULTS

1,930 workers had complete data for analysis. The models underestimated the risk of high SA in eldercare workers and the SA episodes model had to be re-calibrated to the Danish data. Discrimination was practically useful for the re-calibrated SA episodes model, but not the SA days model. Physical workload improved the SA days model (IDI = 0.40; 95% CI 0.19-0.60) and psychosocial work factors, particularly the quality of leadership (IDI = 0.70; 95% CI 053-0.86) improved the SA episodes model.

CONCLUSIONS

The prognostic model predicting high SA days showed poor performance even after physical workload was added. The prognostic model predicting high SA episodes could be used to identify high-risk workers, especially when psychosocial work factors are added as predictor variables.

Risk reclassification analysis investigating the added value of fatigue to sickness absence predictions

Background

Prognostic models including age, self-rated health and prior sickness absence (SA) have been found to predict high (≥30) SA days and high (≥3) SA episodes during 1-year follow-up. More predictors of high SA are needed to improve these SA prognostic models. The purpose of this study was to investigate fatigue as new predictor in SA prognostic models by using risk reclassification methods and measures.

Methods

This was a prospective cohort study with 1-year follow-up of 1,137 office workers. Fatigue was measured at baseline with the 20-item checklist individual strength and added to the existing SA prognostic models. SA days and episodes during 1-year follow-up were retrieved from an occupational health service register. The added value of fatigue was investigated with Net Reclassification Index (NRI) and integrated discrimination improvement (IDI) measures.

Results

In total, 579 (51 %) office workers had complete data for analysis. Fatigue was prospectively associated with both high SA days and episodes. The NRI revealed that adding fatigue to the SA days model correctly reclassified workers with high SA days, but incorrectly reclassified workers without high SA days. The IDI indicated no improvement in risk discrimination by the SA days model. Both NRI and IDI showed that the prognostic model predicting high SA episodes did not improve when fatigue was added as predictor variable.

Conclusion

In the present study, fatigue increased false-positive rates which may reduce the cost-effectiveness of interventions for preventing SA.

Optimal cutoff values of WHO-HPQ presenteeism scores by ROC analysis for preventing mental sickness absence in Japanese prospective cohort

Objectives

Sickness absence due to mental disease in the workplace has become a global public health problem. Previous studies report that sickness presenteeism is associated with sickness absence. We aimed to determine optimal cutoff scores for presenteeism in the screening of the future absences due to mental disease.

Methods

A prospective study of 2195 Japanese employees from all areas of Japan was conducted. Presenteeism and depression were measured by the validated Japanese version of the World Health Organization Health and Work Performance Questionnaire (WHO-HPQ) and K6 scale, respectively. Absence due to mental disease across a 2-year follow-up was surveyed using medical certificates obtained for work absence. Socioeconomic status was measured via a self-administered questionnaire. Receiver operating curve (ROC) analysis was used to determine optimal cutoff scores for absolute and relative presenteeism in relation to the area under the curve (AUC), sensitivity, and specificity.

Results

The AUC values for absolute and relative presenteeism were 0.708 (95% CI, 0.618–0.797) and 0.646 (95% CI, 0.546–0.746), respectively. Optimal cutoff scores of absolute and relative presenteeism were 40 and 0.8, respectively. With multivariate adjustment, cohort participants with our proposal cutoff scores for absolute and relative presenteeism were significantly more likely to be absent due to mental disease (OR = 4.85, 95% CI: 2.20–10.73 and OR = 5.37, 95% CI: 2.42–11.93, respectively). The inclusion or exclusion of depressive symptoms (K6≥13) at baseline in the multivariate adjustment did not influence the results.

Conclusions

Our proposed optimal cutoff scores of absolute and relative presenteeism are 40 and 0.8, respectively. Participants who scored worse than the cutoff scores for presenteeism were significantly more likely to be absent in future because of mental disease. Our findings suggest that the utility of presenteeism in the screening of sickness absence due to mental disease would help prevent such an absence.