The individual over time: time series applications in health care research

Time Series Prediction of Lung Cancer Death Rates on the Basis of SEER Data

PURPOSE

The purpose of this study was to apply different time series analytical techniques to SEER US lung cancer death rate data to develop a best fit model.

METHODS

Three models for yearly time series predictions were built: autoregressive integrated moving average (ARIMA), simple exponential smoothing (SES), and Holt's double expansional smoothing (HDES) models. The three models were built using Python 3.9, on the basis of Anaconda 2022.10.

RESULTS

This study used SEER data from 1975 to 2018 and included 545,486 patients with lung cancer. The best parameters for ARIMA are ARIMA (p, d, q) = (0, 2, 2). In addition, the best parameter for SES was α = .995, whereas the best parameters for HDES were α = .4 and β = .9. The HDES was the model that best fit the lung cancer death rate data, with a root mean square error (RMSE) of 132.91.

CONCLUSION

Including monthly diagnoses, death rates, and years in SEER data increases the number of observations for training and test sets, enhancing the performance of time series models. The reliability of the RMSE was based on the mean lung cancer mortality rate. Owing to the high mean lung cancer death rate of 8,405 patients per year, it is acceptable for reliable models to have large RMSEs.

Incomplete time-series gene expression in integrative study for islet autoimmunity prediction

Type 1 diabetes (T1D) outcome prediction plays a vital role in identifying novel risk factors, ensuring early patient care and designing cohort studies. TEDDY is a longitudinal cohort study that collects a vast amount of multi-omics and clinical data from its participants to explore the progression and markers of T1D. However, missing data in the omics profiles make the outcome prediction a difficult task. TEDDY collected time series gene expression for less than 6% of enrolled participants. Additionally, for the participants whose gene expressions are collected, 79% time steps are missing. This study introduces an advanced bioinformatics framework for gene expression imputation and islet autoimmunity (IA) prediction. The imputation model generates synthetic data for participants with partially or entirely missing gene expression. The prediction model integrates the synthetic gene expression with other risk factors to achieve better predictive performance. Comprehensive experiments on TEDDY datasets show that: (1) Our pipeline can effectively integrate synthetic gene expression with family history, HLA genotype and SNPs to better predict IA status at 2 years (sensitivity 0.622, AUC 0.715) compared with the individual datasets and state-of-the-art results in the literature (AUC 0.682). (2) The synthetic gene expression contains predictive signals as strong as the true gene expression, reducing reliance on expensive and long-term longitudinal data collection. (3) Time series gene expression is crucial to the proposed improvement and shows significantly better predictive ability than cross-sectional gene expression. (4) Our pipeline is robust to limited data availability. Availability: Code is available at https://github.com/compbiolabucf/TEDDY.

Multi-centre implementation of an Educational program to improve the Cardiac Arrest diagnostic accuracy of ambulance Telecommunicators and survival outcomes for sudden cardiac arrest victims: the EduCATe study design and methodology

Background

Sudden cardiac death remains a leading cause of mortality in Canada, resulting in more than 35,000 deaths annually. Most cardiac arrest victims collapse in their own home (85% of the time) and 50% are witnessed by a family member or bystander. Survivors have a quality of life similar to the general population, but the overall survival rate for out-of-hospital cardiac arrest (OHCA) rarely exceeds 8%. Victims are almost four times more likely to survive when receiving bystander CPR, but bystander CPR rates have remained low in Canada over the past decade, not exceeding 15–25% until recently. Telecommunication-assisted CPR instructions have been shown to significantly increase bystander CPR rates, but agonal breathing may be misinterpreted as a sign of life by 9–1-1 callers and telecommunicators, and is responsible for as much as 50% of missed OHCA diagnoses.

We sought to improve the ability and speed with which ambulance telecommunicators can recognize OHCA over the phone, initiate timely CPR instructions, and improve survival.

Methods

In this multi-center national study, we will implement and evaluate an educational program developed for ambulance telecommunicators using a multiple baseline interrupted time-series design. We will compare outcomes 12 months before and after the implementation of a 20-min theory-based educational video addressing barriers to recognition of OHCA while in the presence of agonal breathing. Participating Canadian sites demonstrated prior ability to collect standardized data on OHCA. Data will be collected from eligible 9–1-1 recordings, paramedic documentation and hospital medical records. Eligible cases will include suspected or confirmed OHCA of presumed cardiac origin in patients of any age with attempted resuscitation.

Discussion

The ability of telecommunication-assisted CPR instructions to improve bystander CPR and survival rates for OHCA victims is undeniable. The ability of telecommunicators to recognize OHCA over the phone is unequivocally impeded by relative lack of training on agonal breathing, and reluctance to initiate CPR instructions when in doubt. Our pilot data suggests the potential impact of this project will be to increase absolute OHCA recognition and bystander CPR rates by at least 10%, and absolute out-of-hospital cardiac arrest survival by 5% or more.

Trial registration

Prospectively registered on March 28, 2019 at ClinicalTrials.gov identifier: NCT03894059.

Three Statistical Approaches for Assessment of Intervention Effects: A Primer for Practitioners

INTRODUCTION

Statistical methods to assess the impact of an intervention are increasingly used in clinical research settings. However, a comprehensive review of the methods geared toward practitioners is not yet available.

METHODS AND MATERIALS

We provide a comprehensive review of three methods to assess the impact of an intervention: difference-in-differences (DID), segmented regression of interrupted time series (ITS), and interventional autoregressive integrated moving average (ARIMA). We also compare the methods, and provide illustration of their use through three important healthcare-related applications.

RESULTS

In the first example, the DID estimate of the difference in health insurance coverage rates between expanded states and unexpanded states in the post-Medicaid expansion period compared to the pre-expansion period was 5.93 (95% CI, 3.99 to 7.89) percentage points. In the second example, a comparative segmented regression of ITS analysis showed that the mean imaging order appropriateness score in the emergency department at a tertiary care hospital exceeded that of the inpatient setting with a level change difference of 0.63 (95% CI, 0.53 to 0.73) and a trend change difference of 0.02 (95% CI, 0.01 to 0.03) after the introduction of a clinical decision support tool. In the third example, the results from an interventional ARIMA analysis show that numbers of creatinine clearance tests decreased significantly within months of the start of eGFR reporting, with a magnitude of drop equal to -0.93 (95% CI, -1.22 to -0.64) tests per 100,000 adults and a rate of drop equal to 0.97 (95% CI, 0.95 to 0.99) tests per 100,000 per adults per month.

DISCUSSION

When choosing the appropriate method to model the intervention effect, it is necessary to consider the structure of the data, the study design, availability of an appropriate comparison group, sample size requirements, whether other interventions occur during the study window, and patterns in the data.

Bias in Before-After Studies: Narrative Overview for Anesthesiologists

Before-after study designs are effective research tools and in some cases, have changed practice. These designs, however, are inherently susceptible to bias (ie, systematic errors) that are sometimes subtle but can invalidate their conclusions. This overview provides examples of before-after studies relevant to anesthesiologists to illustrate potential sources of bias, including selection/assignment, history, regression to the mean, test-retest, maturation, observer, retrospective, Hawthorne, instrumentation, attrition, and reporting/publication bias. Mitigating strategies include using a control group, blinding, matching before and after cohorts, minimizing the time lag between cohorts, using prospective data collection with consistent measuring/reporting criteria, time series data collection, and/or alternative study designs, when possible. Improved reporting with enforcement of the Enhancing Quality and Transparency of Health Research (EQUATOR) checklists will serve to increase transparency and aid in interpretation. By highlighting the potential types of bias and strategies to improve transparency and mitigate flaws, this overview aims to better equip anesthesiologists in designing and/or critically appraising before-after studies.

FuseAD: Unsupervised Anomaly Detection in Streaming Sensors Data by Fusing Statistical and Deep Learning Models

The need for robust unsupervised anomaly detection in streaming data is increasing rapidly in the current era of smart devices, where enormous data are gathered from numerous sensors. These sensors record the internal state of a machine, the external environment, and the interaction of machines with other machines and humans. It is of prime importance to leverage this information in order to minimize downtime of machines, or even avoid downtime completely by constant monitoring. Since each device generates a different type of streaming data, it is normally the case that a specific kind of anomaly detection technique performs better than the others depending on the data type. For some types of data and use-cases, statistical anomaly detection techniques work better, whereas for others, deep learning-based techniques are preferred. In this paper, we present a novel anomaly detection technique, FuseAD, which takes advantage of both statistical and deep-learning-based approaches by fusing them together in a residual fashion. The obtained results show an increase in area under the curve (AUC) as compared to state-of-the-art anomaly detection methods when FuseAD is tested on a publicly available dataset (Yahoo Webscope benchmark). The obtained results advocate that this fusion-based technique can obtain the best of both worlds by combining their strengths and complementing their weaknesses. We also perform an ablation study to quantify the contribution of the individual components in FuseAD, i.e., the statistical ARIMA model as well as the deep-learning-based convolutional neural network (CNN) model.

The impact of randomized trial results on abdominal aortic aneurysm repair rates from 2003 to 2016: A population-based time-series analysis

Objectives

The uptake of endovascular aortic repair for elective and ruptured abdominal aortic aneurysm repair is not well studied. We aimed to examine the trends in open surgical repair and endovascular aortic repair of eAAA and rAAA and to examine the effects of randomized trial publications on elective open surgical repair and endovascular aortic repair rates.

Methods

We conducted a population-based time-series analysis of eAAA and rAAA repairs in Ontario, Canada from 2003 to 2016. We examined changes in overall and approach-specific rates of eAAA and rAAA repair using exponential smoothing models. Interventional autoregressive integrated moving average models were fit to the eAAA rates to examine the impact of randomized trial results on these rates.

Results

We identified 19,489 eAAA (12,232 open (63%) and 7257 endovascular (37%)) and 2732 rAAA (2466 open (90%) and 266 endovascular (10%)) repairs from 2003 to 2016. The rate of eAAA repair declined from 6.39/100,000 in 2003 to 5.59/100,000 in 2016 (13% decrease, p = 0.17). The rate of elective open surgical repair decreased nearly three-fold from 6.07/100,000 to 2.12/100,000 ( p < 0.0001), while elective endovascular aortic repair increased approximately 10-fold (0.32/100,000 to 3.47/100,000, p < 0.0001). The rate of ruptured open surgical repair decreased from 1.62/100,000 to 0.37/100,000 ( p < 0.44), while ruptured endovascular aortic repair uptake increased (0.00/100,000 to 0.12/100,000, p < 0.25). The mid-term results of the DREAM and EVAR-1 trials were associated with a decrease in the rate of elective open surgical repair decline after 2010 ( p = 0.01).

Conclusions

While elective open surgical repair use has significantly decreased from 2003 to 2016, elective endovascular aortic repair use has significantly increased. The DREAM and EVAR-1 results significantly impacted the observed rates of elective open surgical repair only. The reasons for these trends require further characterization.

Prevalence of Elective and Ruptured Abdominal Aortic Aneurysm Repairs by Age and Sex From 2003 to 2016 in Ontario, Canada

IMPORTANCE

Age and sex are important considerations in assessing and individualizing therapy for abdominal aortic aneurysm (AAA) repair.

OBJECTIVE

To determine the prevalence of open and endovascular elective AAA (EAAA) and ruptured AAA (RAAA) repair by age and sex.

DESIGN, SETTING, AND PARTICIPANTS

In this retrospective, population-based, cross-sectional, time-series analysis in Ontario, Canada, from April 1, 2003, to March 31, 2016, all patients undergoing AAA repair who were older than 39 years were included.

EXPOSURES

Elective AAA and RAAA repair with open surgical repair (OSR) or endovascular aortic repair (EVAR).

MAIN OUTCOMES AND MEASURES

Age- and sex-standardized rates of EAAA and RAAA repair with OSR and EVAR.

RESULTS

From 2003 to 2016, 19 489 EAAA repairs (12 232 [63%] OSR and 7257 [37%] EVAR) and 2732 RAAA repairs (2466 [90%] OSR and 266 [10%] EVAR) were identified. The mean (SD) age was 72.7 (8.1) years in the EAAA subgroup and 73.5 (8.9) years in the RAAA subgroup; 15 813 patients (81%) in the EAAA subgroup and 2178 (80%) in the RAAA subgroup were men. The rates of EAAA by age quintile and sex decreased over the study period except among patients older than 79 years (1.3 per 100 000 population in 2003 to 2.2 per 100 000 population in 2016; 70% increase; P < .001). The rates of elective OSR decreased across all age and sex subgroups (range, 38%-74% decrease; P ≤ .009 for all subgroups) except among patients older than 79 years (1.3 per 100 000 population at baseline to 0.56 per 100 000 population in the second quarter of 2016; 53% decrease; P = .05). The rates of elective EVAR significantly increased across all age and sex subgroups (range, 566%-1585% increase; P ≤ .04 for all subgroups). Elective EVAR became the dominant treatment approach for aneurysms in men around 2010, whereas it maintained parity among women in 2016. The RAAA repair rate decreased over the study period in all subgroups (range, 32%-91% decrease; P ≤ .001 for all subgroups), but the decrease was not significant among women (80% decrease; P = .08). Similarly, the rates of ruptured OSR decreased among all subgroups (range, 47%-91% decrease; P < .001), but the decrease was not significant among women (87% decrease; P = .54). Ruptured EVAR showed significant uptake in all subgroups.

CONCLUSIONS AND RELEVANCE

Among patients with AAA in Ontario, Canada, use of EVAR appeared to increase from 2003 to 2016, whereas OSR use appeared to decrease. These findings were most pronounced among elective procedures for men and older patients. The delayed increase in the use of EVAR among women may reflect continued anatomical constraints for women seeking elective repair.

Semi-Automated Visualization and ANalysis of Trends: A "SAVANT" for Facilitating Antimicrobial Stewardship Using Antistaphylococcal Resistance and Consumption as a Prototype

Background

Governments and health care regulators now require hospitals and nursing homes to establish programs to monitor and report antimicrobial consumption and resistance. However, additional resources were not provided. We sought to develop an approach for monitoring antimicrobial resistance and consumption that health care systems can implement with minimal added costs or modifications to existing diagnostic and informatics infrastructure.

Methods

Using (1) the electronic laboratory information system of a nationwide managed care network, (2) the 3 most widely used commercial microbiology diagnostic platforms, and (3) Staphylococcus aureus, one of the most common causes of infections worldwide, as a prototype, we validated the approach dubbed "SAVANT" for Semi-Automated Visualization and ANalysis of Trends. SAVANT leverages 3 analytical methods (time series analysis, the autoregressive integrated moving average, and generalized linear regression) on either commercial or open source software to report trends in antistaphylococcal use and resistance.

Results

All laboratory results from January 2010 through December 2015 from an annual average of 9.2 million health care beneficiaries were queried. Inpatient and outpatient prescription rates were calculated for 8 key antistaphylococcal compounds. Trends and relationships of antistaphylococcal consumption and resistance among 81 840 unique S. aureus isolates from >6.5 million cultures were revealed.

Conclusions

Using existing or freely available resources, SAVANT was successfully implemented across a complex and geographically dispersed 280-hospital network, bridging a critical gap between medical informatics, large-scale data analytics, and mandatory reporting of health care quality metrics.

Magic Mirror, on the Wall-Which Is the Right Study Design of Them All?-Part I

The assessment of a new or existing treatment or intervention typically answers 1 of 3 research-related questions: (1) "Can it work?" (efficacy); (2) "Does it work?" (effectiveness); and (3) "Is it worth it?" (efficiency or cost-effectiveness). There are a number of study designs that on a situational basis are appropriate to apply in conducting research. These study designs are classified as experimental, quasi-experimental, or observational, with observational studies being further divided into descriptive and analytic categories. This first of a 2-part statistical tutorial reviews these 3 salient research questions and describes a subset of the most common types of experimental and quasi-experimental study design. Attention is focused on the strengths and weaknesses of each study design to assist in choosing which is appropriate for a given study objective and hypothesis as well as the particular study setting and available resources and data. Specific studies and papers are highlighted as examples of a well-chosen, clearly stated, and properly executed study design type.

Impact of Clinical Trial Results on the Temporal Trends of Carotid Endarterectomy and Stenting From 2002 to 2014

Supplemental Digital Content is available in the text.

Background and Purpose—

Randomized trials provide conflicting data for the efficacy of carotid-artery stenting compared with endarterectomy. The purpose of this study was to examine the impact of conflicting clinical trial publications on the utilization rates of carotid revascularization procedures.

Methods—

We conducted a population-level time-series analysis of all individuals who underwent carotid endarterectomy and stenting in Ontario, Canada (2002–2014). The primary analysis examined temporal changes in the rates of carotid revascularization procedures after publications of major randomized trials. Secondary analyses examined changes in overall and age, sex, carotid-artery symptom, and operator specialty–specific procedure rates.

Results—

A total of 16 772 patients were studied (14 394 endarterectomy [86%]; 2378 stenting [14%]). The overall rate of carotid revascularization decreased from 6.0 procedures per 100 000 individuals ≥40 years old in April 2002 to 4.3 procedures in the first quarter of 2014 (29% decrease; P<0.001). The rate of endarterectomy decreased by 36% (P<0.001), whereas the rate of carotid-artery stenting increased by 72% (P=0.006). We observed a marked increase (P=0.01) in stenting after publication of the SAPPHIRE trial (Stenting and Angioplasty With Protection in Patients at High Risk for Endarterectomy) in 2004, whereas stenting remained relatively unchanged after subsequent randomized trials published in 2006 (P=0.11) and 2010 (P=0.34). In contrast, endarterectomy decreased after trials published in 2006 (P=0.04) and 2010 (P=0.005).

Conclusions—

Although the overall rates of carotid revascularization and endarterectomy have fallen since 2002, the rate of carotid-artery stenting has risen since the publication of stenting-favorable SAPPHIRE trial. Subsequent conflicting randomized trials were associated with a decreasing rate of carotid endarterectomy.

Research Methods in Healthcare Epidemiology and Antimicrobial Stewardship-Quasi-Experimental Designs

Quasi-experimental studies evaluate the association between an intervention and an outcome using experiments in which the intervention is not randomly assigned. Quasi-experimental studies are often used to evaluate rapid responses to outbreaks or other patient safety problems requiring prompt, nonrandomized interventions. Quasi-experimental studies can be categorized into 3 major types: interrupted time-series designs, designs with control groups, and designs without control groups. This methods paper highlights key considerations for quasi-experimental studies in healthcare epidemiology and antimicrobial stewardship, including study design and analytic approaches to avoid selection bias and other common pitfalls of quasi-experimental studies. Infect Control Hosp Epidemiol 2016;1-6.

Time-series analysis of continuously monitored blood glucose: the impacts of geographic and daily lifestyle factors

Type 2 diabetes is known to be associated with environmental, behavioral, and lifestyle factors. However, the actual impacts of these factors on blood glucose (BG) variation throughout the day have remained relatively unexplored. Continuous blood glucose monitors combined with human activity tracking technologies afford new opportunities for exploration in a naturalistic setting. Data from a study of 40 patients with diabetes is utilized in this paper, including continuously monitored BG, food/medicine intake, and patient activity/location tracked using global positioning systems over a 4-day period. Standard linear regression and more disaggregated time-series analysis using autoregressive integrated moving average (ARIMA) are used to explore patient BG variation throughout the day and over space. The ARIMA models revealed a wide variety of BG correlating factors related to specific activity types, locations (especially those far from home), and travel modes, although the impacts were highly personal. Traditional variables related to food intake and medications were less often significant. Overall, the time-series analysis revealed considerable patient-by-patient variation in the effects of geographic and daily lifestyle factors. We would suggest that maps of BG spatial variation or an interactive messaging system could provide new tools to engage patients and highlight potential risk factors.

Trend and seasonality in hospitalizations for pulmonary embolism: a time-series analysis

BACKGROUND

The existence of seasonal variability in patients with acute pulmonary embolism (PE) has been debated for years, with contradictory results. The aim of this study was to identify the trend and possible existence of a seasonal pattern in hospitalizations for PE in Spain.

METHODS

We analyzed the hospital discharge database of the Spanish National Health System from 2001 to 2010. Patients aged > 14 years diagnosed with PE were selected and a time series was constructed considering mean daily admissions for PE by month. The trend and seasonality factor of the series were determined using time-series analysis, and time-series modeling was used for analysis. Exponential smoothing models and the autoregressive integrated moving average test were used to generate a predictive model.

RESULTS

From 2001 to 2010, there were 162,032 diagnoses of PE (5.07 per 1000 hospitalizations). In 105,168 cases, PE was the reason for admission. The PE diagnosis rate ranged from 4.14 per 1000 in 2001 to 6.56 per 1000 in 2010; and hospital admissions due to PE ranged from 2.67 to 4.28 per 1000 hospital discharges. Time-series analysis showed a linear increase in the incidence and a significant seasonal pattern with 17% more admissions in February and 12% fewer in June-July with respect to the central tendency (difference from February to June, 29%).

CONCLUSIONS

The incidence of hospitalizations for PE showed a linear increase and a seasonal pattern, with the highest number of admissions in winter and the lowest number in summer.

Forecasting ESKAPE infections through a time-varying auto-adaptive algorithm using laboratory-based surveillance data

Background

Mathematical or statistical tools are capable to provide a valid help to improve surveillance systems for healthcare and non-healthcare-associated bacterial infections. The aim of this work is to evaluate the time-varying auto-adaptive (TVA) algorithm-based use of clinical microbiology laboratory database to forecast medically important drug-resistant bacterial infections.

Methods

Using TVA algorithm, six distinct time series were modelled, each one representing the number of episodes per single ‘ESKAPE’ (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa and Enterobacter species) infecting pathogen, that had occurred monthly between 2002 and 2011 calendar years at the Università Cattolica del Sacro Cuore general hospital.

Results

Monthly moving averaged numbers of observed and forecasted ESKAPE infectious episodes were found to show a complete overlapping of their respective smoothed time series curves. Overall good forecast accuracy was observed, with percentages ranging from 82.14% for E. faecium infections to 90.36% for S. aureus infections.

Conclusions

Our approach may regularly provide physicians with forecasted bacterial infection rates to alert them about the spread of antibiotic-resistant bacterial species, especially when clinical microbiological results of patients’ specimens are delayed.

Electronic supplementary material

The online version of this article (doi:10.1186/s12879-014-0634-9) contains supplementary material, which is available to authorized users.

Menstrual cycle and detectable human papillomavirus in reproductive-age women: a time series study

BACKGROUND

Current evidence on the relationship between human papillomavirus (HPV) DNA detection and menstrual cycle has been inconsistent.

METHODS

We included 21 nonoral contraceptive pill (non-OCP) users who self-collected vaginal samples twice per week for 16 weeks. We explored whether variable detection of HPV DNA exhibited cyclic or other structured temporal patterns. We also evaluated relationships between serial HPV prevalence, sexual behavior, and suspected bacterial vaginosis (BV) as defined by Nugent Gram stain score ≥7.

RESULTS

During follow-up, any-type HPV prevalence varied between 61.1% and 85.0%. Although not statistically significant, we observed a maximum autocorrelation in serial HPV prevalence lagging 14 days (correlation coefficient [ρ], -0.24). Any-type HPV detection had a periodic behavior, generally repeating every 28.0 days (bootstrapped interquartile range, 22.4-28.0) and peaking around the ovulation time (adjusted odds ratio, 1.96; 95% confidence interval [CI], 1.06-3.62) as compared to menstruation. We also showed that an increase in any-type HPV prevalence preceded the beginning of a menstrual cycle by 9-12 days. There was no evidence of relationships between HPV prevalence and sexual activity or Nugent score.

CONCLUSIONS

Serially detected any-type HPV DNA showed a periodic behavior and was likely to peak in the periovulatory phase among non-OCP users.

Conventional and advanced time series estimation: application to the Australian and New Zealand Intensive Care Society (ANZICS) adult patient database, 1993-2006

RATIONALE

Time series analysis has seen limited application in the biomedical Literature. The utility of conventional and advanced time series estimators was explored for intensive care unit (ICU) outcome series.

METHODS

Monthly mean time series, 1993-2006, for hospital mortality, severity-of-illness score (APACHE III), ventilation fraction and patient type (medical and surgical), were generated from the Australia and New Zealand Intensive Care Society adult patient database. Analyses encompassed geographical seasonal mortality patterns, series structural time changes, mortality series volatility using autoregressive moving average and Generalized Autoregressive Conditional Heteroscedasticity models in which predicted variances are updated adaptively, and bivariate and multivariate (vector error correction models) cointegrating relationships between series.

RESULTS

The mortality series exhibited marked seasonality, declining mortality trend and substantial autocorrelation beyond 24 lags. Mortality increased in winter months (July-August); the medical series featured annual cycling, whereas the surgical demonstrated long and short (3-4 months) cycling. Series structural breaks were apparent in January 1995 and December 2002. The covariance stationary first-differenced mortality series was consistent with a seasonal autoregressive moving average process; the observed conditional-variance volatility (1993-1995) and residual Autoregressive Conditional Heteroscedasticity effects entailed a Generalized Autoregressive Conditional Heteroscedasticity model, preferred by information criterion and mean model forecast performance. Bivariate cointegration, indicating long-term equilibrium relationships, was established between mortality and severity-of-illness scores at the database level and for categories of ICUs. Multivariate cointegration was demonstrated for {log APACHE III score, log ICU length of stay, ICU mortality and ventilation fraction}.

CONCLUSIONS

A system approach to understanding series time-dependence may be established using conventional and advanced econometric time series estimators.

Changes in weather and the effects on pediatric asthma exacerbations

BACKGROUND

Pediatric asthma exacerbations may correlate with changes in weather, yet this relationship is not well defined.

OBJECTIVE

To determine the effects of fluctuations in climatic factors (temperature, humidity, and barometric pressure) on pediatric asthma exacerbations.

METHODS

A retrospective study was performed at 1 large urban hospital during a 2-year period (January 1, 2004, to December 31, 2005). Children presenting to the emergency department (ED) for an asthma exacerbation were included. Data on climactic factors, pollutants, and aeroallergens were collected daily. The relationship of daily (intraday) or between-day (interday) changes in climactic factors and asthma ED visits was evaluated using time series analysis, controlling for seasonality, air pollution, and aeroallergen exposure. The effects of climactic factors were evaluated on the day of admission (T=0) and up to 5 days before admission (T-5 through T-1).

RESULTS

There were 25,401 asthma ED visits. A 10% intraday increase in humidity on day T-1 or day T-2 was associated with approximately 1 additional ED visit for asthma (P < .001 and P = .01, respectively). Interday changes in humidity from day T - 3 to T-2 were also associated with more ED visits (P < .001). Interday changes in temperature from T-1 to T = 0 increased ED visits, with a 10 degrees F increase being associated with 1.8 additional visits (P = .006). No association was found with changes in barometric pressure.

CONCLUSION

Fluctuations in humidity and temperature, but not barometric pressure, appear to influence ED visits for pediatric asthma. The additional ED visits occur 1 to 2 days after the fluctuation.

Randomised, controlled, cross-over trial of soy protein with isoflavones on blood pressure and arterial function in hypertensive subjects

OBJECTIVE

To examine the effects of dietary soy/isoflavones on 24 hr blood pressure profiles and arterial function [systemic arterial compliance (SAC), pulse wave velocity (PWV) and brachial arterial flow mediated vasodilation (FMD)] compared to non legume-based plant protein without isoflavones, in hypertensive subjects.

DESIGN

In a 6 month double-blind, placebo controlled, cross-over trial, 41 hypertensive subjects (26 men, 15 postmenopausal women), 30-75 years, received soy cereal (40 g soy protein, 118 mg isoflavones) and gluten placebo cereal, each for 3 months.

RESULTS

Thirty-eight subjects completed protocol with results expressed as mean or mean change (+/-SEM) with each intervention. Soy increased urinary isoflavones (daidzein: 8-fold; genistein: 8-fold; equol: 9-fold; ODMA: 18-fold) with no change during gluten placebo. There was no difference in the change in individual 24 hr ambulatory BP parameters (SBP: 2 +/- 2 vs -1 +/- 1 mmHg, p = 0.21; DBP: 1 +/- 1 vs -1 +/- 1 mmHg, p = 0.06) central BP (cSBP: -4 +/- 2 vs 0 +/- 2 mmHg, p = 0.2) or the change in arterial function (FMD: 0.3 +/- 0.5 vs -0.2 +/- 0.5%, p = NS; SAC: 0.02 +/- 0.02 vs -0.02 +/- 0.02 U/mmHg, p = NS; PWV central: -0.2 +/- 0.2 vs 0.0 +/- 0.2 m/sec, p = NS; PWV peripheral: 0.01 +/- 0.3 vs -0.4 +/- 0.4 m/sec, p = NS) noted between interventions. Analysis of the area under curve of 24 hr BP outputs demonstrated that soy protein compared to gluten protein resulted in higher 24 hr systolic BP by 2.3 mmHg (p = 0.003), a higher daytime systolic BP by 3.4 mmHg (p = 0.0002) and a higher daytime diastolic BP by 1.4 mmHg (p = 0.008). Overall 24 hr diastolic BP, night systolic BP and night diastolic BP were not significantly different between groups. Furthermore, soy protein compared to gluten protein resulted in higher 24 hr heart rates by 3.5 bpm (p < 0.0001).

CONCLUSIONS

In hypertensive subjects, compared to gluten placebo, soy dietary supplementation containing isoflavones had no effect on arterial function, on average 24 hr ambulatory blood pressure parameters or central blood pressure in men and women with hypertension. Area under the curve of 24 hr profiles demonstrated that daytime BP was higher after soy compared to gluten.

The examination of headache activity using time-series research designs

The majority of research conducted on headache has utilized cross-sectional designs which preclude the examination of dynamic factors and principally rely on group-level effects. The present article describes the application of an individual-oriented process model using time-series analytical techniques. The blending of a time-series approach with an interactive process model allows consideration of the relationships of intra-individual dynamic processes, while not precluding the researcher to examine inter-individual differences. The authors explore the nature of time-series data and present two necessary assumptions underlying the time-series approach. The concept of shock and its contribution to headache activity is also presented. The time-series approach is not without its problems and two such problems are specifically reported: autocorrelation and the distribution of daily observations. The article concludes with the presentation of several analytical techniques suited to examine the time-series interactive process model.

Changes in mortality after the recent economic crisis in South Korea

PURPOSE

To examine the changes in all cause mortality and cause-specific mortality after the economic crisis in South Korea.

METHODS

Monthly mortality data for an entire country was used and intervention analysis applied to compare mortality after the crisis with mortality which would have occurred if the trends before the crisis had continued.

RESULTS

All cause mortality began to increase about 1 year after the crisis, while cardiovascular increased immediately. Transport accidents decreased significantly during the year following the crisis and then regressed towards the pre-economic crisis level. Suicides increased rapidly and maintained an upward trend but subsequently reduced towards the pre-economic crisis level.

CONCLUSIONS

This study has shown an evidence of a relationship between economic crisis and mortality.

The impact of guidelines on mild hypertension in pregnancy: time series analysis

OBJECTIVE

To determine the impact of a national strategy to promote implementation of a guideline on the management of mild, non-proteinuric hypertension in pregnancy.

DESIGN

Simple, interrupted time series analysis.

SETTING

Four maternity units in Scotland.

POPULATION

Women delivering a live or stillborn baby.

METHODS

Dissemination of the guideline under the auspices of a national clinical effectiveness programme, supported by a national launch meeting and feedback from a survey of obstetricians highlighting aspects of care that could be improved.

MAIN OUTCOME MEASURES

Appropriateness of initial investigation and subsequent clinical management, and costs of guideline development and implementation activities.

DATA COLLECTION

Twenty-four months pre-intervention and 12 months post-intervention data were abstracted from a random sample of case notes.

RESULTS

Initial investigation was consistent with recommendations for 59.9% out of 1263 women and subsequent clinical management for 67.6% out of 1081 in whom a diagnosis could be made from available data. There were no significant changes in the appropriateness of initial investigation (10.6%; 95% confidence interval [CI] -0.1% to 19.3%; decreasing by 1.2% per month post-implementation, 95% CI -2.5% to 0.1%) or clinical management (-0.3%; 95% CI -8.7% to 11.2%). Guideline development and implementation cost an estimated pound 2784 per maternity unit in Scotland.

CONCLUSIONS

Clinical care of mild hypertension in pregnancy remains highly inconsistent. The lack of the intervention effect may be related to the complexity of the guideline recommendations and the nature of the implementation strategy.

Interrupted time series analysis in clinical research

OBJECTIVE

To demonstrate the usefulness of interrupted time series analysis in clinical trial design.

METHODS

A safety data set of electrocardiographic (ECG) information was simulated from actual data that had been collected in a Phase I study. Simulated data on 18 healthy volunteers based on a study performed in a contract research facility were collected based on single doses of an experimental medication that may affect ECG parameters. Serial ECGs were collected before and during treatment with the experimental medication. Data from 7 real subjects receiving placebo were used to simulate the pretreatment phase of time series; data from 18 real subjects receiving active treatment were used to simulate the treatment phase of the time series. Visual inspection of data was performed, followed by tests for trend, seasonality, and autocorrelation by use of SAS.

RESULTS

There was no evidence of trend, seasonality, or autocorrelation. In 11 of 18 simulated individuals, statistically significant changes in QTc intervals were observed following treatment with the experimental medication. A significant time of day and treatment interaction was observed in 4 simulated patients.

CONCLUSIONS

Interrupted time series analysis techniques offer an additional tool for the study of clinical situations in which patients must act as their own controls and where serial data can be collected at evenly distributed intervals.

Stimulating Adverse Drug Reaction Reporting

BACKGROUND

The effectiveness of voluntary reporting systems in pharmacovigilance highly depends on the number of assembled reports.

AIM

The aim of this study was to measure the effect of the periodical distribution of a bulletin on drug safety issues and of including yellow cards in prescription pads on the rate of adverse drug reaction (ADR) reporting.

STUDY DESIGN AND METHODS

The Catalan Centre of Pharmacovigilance began its activities at the end of 1982. Since 1985, an ADR bulletin (ADRB) has been mailed approximately quarterly to all physicians in its catchment area, with one yellow card enclosed. Additionally, from 1991-1994, a yellow card was included in the prescription pads of the Catalan Health Service. Time series methodology, with adjustment of the monthly number of reports to an Auto-Regressive Integrated Moving Average (ARIMA) model, was used to evaluate the effect of these two measures.

RESULTS

From January 1983-October 1995, 6240 spontaneous ADR reports were received, and 41 issues of the ADRB were sent out. Initially, the mean monthly spontaneous ADR reporting rate was 34.4 (SD = 14.1; n = 106 months). After the inclusion of yellow cards in prescription pads, the mean monthly spontaneous ADR reporting rate increased to 53.9 (SD = 14.4; n = 48 months). According to an ARIMA model, when a bulletin was send out (MONTH1), a mean increase of 9.4 reports was produced in that month, plus 12.3 additional reports in the following month (MONTH2), and 6.3 in the second month after sending the ADRB (MONTH3). A yellow card in the prescription pads elicits a monthly mean increase of 19.8 in the number of reports.

CONCLUSIONS

The present study suggests that ADRBs elicit a temporal increase of the ADR reporting rate. Including a yellow card in prescription pads was followed by an even greater increase in the reporting rate, possibly because it guarantees that yellow cards are available at the workplace.

[Meningococcal disease in Spain: seasonal nature and recent changes]

BACKGROUND

In this study we built a time series model in the time domain based on the cases of meningococcal disease declared every four weeks in Spain between 1972 and 1998 With the model we analyzed the incidence of the disease and examined whether an epidemic really took place at the beginning of 1997, or whether the values were in the range of expectations. At this time there was widespread social alarm in Spain, because there was considered to be an outbreak of meningitis. We also examined whether the incidence of the disease diminished during 1997 following the measures introduced by the Comisión Interterritorial del Sistema Nacional de Salud.

METHOD

The data on the incidence of the meningococcal disease in Spain were obtained from the Boletín Epidemiológico Semanal, published by the Instituto de Salud Carlos III, declared from 1972 to 1998 and grouped every four weeks. The method corresponds to the time series model in the domain of time: the Box-Jenkins method.

RESULTS

In order to explain the behaviour of the disease during the years of the study we chose a model ARIMA(0,1,1) (0,1,1)13 that shows annual seasonal behaviour, with higher incidence in the winter and the beginning of spring, and lower incidence in summer. An intervention analysis showed that in March 1996 there was a level shift in the incidence of the illness. This lasted until September 1997, at which time the incidence returned to pre-1996.

CONCLUSIONS

The measures of sanitary monitoring against the meningitis in the months of greater incidence, winter and beginning of spring must be intensified. Between spring 1996 and autumn 1997 the incidence of meningitis increased. Thereafter, until the end of 1998, there was a significant decrease in incidence, attributable to the performance of the health authorities.

A statistical analysis of the seasonality in pulmonary tuberculosis

The present study examines whether pulmonary tuberculosis (PTB) has an annual seasonal pattern. A mathematical model is also obtained to forecast the pattern of incidence. The data for the study are the cases of PTB reported throughout Spain, published in the Epidemiology Bulletin by the Carlos III Health Center of the Spanish Ministry of Health in a 26-year period, 1971-1996. The analytical results show that the low rates in tuberculosis notifications over the period 1971-1981 have changed, halting in 1982 and reversing with high incidence from 1983 onwards. An annual seasonal pattern was also shown with higher incidence during summer and autumn. With the mathematical model we predicted the disease behaviour in 1997 and the results were compared to the reported cases. In Spain, as in several industrialised countries, the reason for this recent increase in the number of reported cases is, mainly, the human immunodeficiency virus (HIV) infection. The seasonal trend, with higher incidence in winter, can be attributed to the increase in indoor activities, much more common than in a warm climate. The tubercle bacilli expelled from infected persons in a room with closed windows may remain infectious for a long time, increasing the risk of exposure of healthy persons to the bacilli. As the preclinical period, from exposure to clinical onset, may be of several weeks, the high incidence in spring would be explained. Moreover, in winter and spring the infections of viral aetiology, like flu, are more frequent and cause immunological deficiency which is another reason for the seasonal trend observed. An incidence greater than that foreseen by the mathematical model would express a failure in epidemiologic surveillance, and thus the results of this study may be used to assess a quality of the preventive measures.

Modelling and forecasting antimicrobial resistance and its dynamic relationship to antimicrobial use: a time series analysis

To investigate the relationship between antimicrobial use and resistance in our hospital, we collected antimicrobial susceptibility and use data from existing microbiology laboratory and pharmacy databases for the period July 1st, 1991-December 31, 1998. The data was analyzed as time series and autoregressive integrated moving average (Box-Jenkins) and transfer function models were built. By using this method, we were able to demonstrate a temporal relationship between antimicrobial use and resistance, to quantify the effect of use on resistance and to estimate the delay between variations of use and subsequent variations in resistance. The results obtained for two antimicrobial-microorganism combinations: ceftazidime-gram-negative bacilli and imipenem-Pseudomonas aeruginosa, are shown as examples.

Box-Jenkins modelling in medical research

Notifications of diseases, entries in a hospital, injuries due to accidents, etc., are frequently collected in fixed equally spaced intervals. Such observations are likely to be dependent. In environmental medicine, where series such as daily concentrations of pollutants are collected and analysed, it is evident that dependence of consecutive measurements may be important. A high concentration of a pollutant today has a certain 'inertia', i.e. a tendency to be high tomorrow as well. Dependence of consecutive observations may be equally important when data such as blood glucose are recorded within a single patient. ARIMA models (autoregressive integrated moving average models, Box-Jenkins models), which allow the stochastic dependence of consecutive data to be modelled, have become well established in such fields as economics. This article reviews basic concepts of Box-Jenkins modelling. The methods are illustrated by applications. In particular, the following topics are presented: the ARIMA model, transfer function models (assessment of relations between time series) and intervention analysis (assessment of changes of time series).

MZ twinning: chance or determinism? An essay in nonlinear dynamics (chaos)

Classically, researchers considered monozygotic twinning (MZT) a random phenomenon. This paper tests the hypothesis with the aid of nonlinear dynamics techniques. The latter can tell true randomness from chance-like variation. Chaos, the endpoint of the threshold state of a nonlinear deterministic system, can mimic constrained randomness. From a practical standpoint, recognizing chaos in a time series data set means that the paradigmatic multifactorial model of causation is essentially ruled out. Specifically, time series of MZ, DZ, and single maternities were analysed. First, spectral analysis was used to uncover periodicities embedded in the series. Second, a singular value decomposition was undertaken to reduce noise from the series. Third, phase space attractors were drawn up that describe the 'asymptotic' trajectory of the system at any time. Results suggested that DZ, MZ, and single maternities shared a similar 32-year periodicity. Owing to two interwoven similar periodicities, the single-maternity cycle kinetics proved to be faster than that of DZ's. The MZ series was the only one to display secondary interacting harmonics, thus eliciting a rather unusual trajectory in the bidimensional phase space. The MZ time points were not spread in a haphazard fashion; on the contrary, a fine structure was present that did not reduce to a limit cycle such as the one characterizing the DZ- or the single-maternity trajectory. It was concluded that a complex nonlinear dynamic underlies MZ twinning. Therefore, calling for extrinsic causes to account for what appears to be random variation overtime would be pointless. MZ twinning should rather be traced to a limited number of intrinsic and deterministic interacting system components. The most likely candidates are presented and discussed.

Forecasting areawide hospital utilization: a comparison of five univariate time series techniques

Time series analysis is one of the methods health services researchers, managers and planners have to examine and predict utilization over time. The focus of this study is univariate time series techniques, which model the change in a dependent variable over time, using time as the only independent variable. These techniques can be used with administrative healthcare databases, which typically contain reliable, time-specific utilization variables, but may lack adequate numbers of variables needed for behavioral or economic modeling. The inpatient discharge database of the Department of Veterans Affairs, the Patient Treatment File, was used to calculate monthly time series over a six-year period for the nation and across US Census Bureau regions for three hospital utilization indicators: average length of stay, discharge rate, and multiple stay ratio, a measure of readmissions. The first purpose of this study was to determine the accuracy of forecasting these indicators 24 months into the future using five univariate time series techniques. In almost all cases, techniques were able to forecast the magnitude and direction of future utilization within a 10% mean monthly error. The second purpose of the study was to describe time series of the three hospital utilization indicators. This approach raised several questions concerning Department of Veterans Affairs hospital utilization.

Time series cross-correlation analysis of postparturient relationships among serum metabolites and yield variables in Holstein cows

Time series cross-correlation analysis was used to determine relationships among serum beta-hydroxybutyrate, glucose, FFA, cholesterol, milk yield, DMI, and estimated net energy balance for 42 d postpartum for 14 multiparous cows. Milk yield was positively associated with beta-hydroxybutyrate concentrations on the same day, and milk yield was a primary determinant of metabolic change. Dry matter intake was negatively correlated with beta-hydroxybutyrate concentrations 3 d later and on the same day. The data supported in vitro studies showing that FFA are positively associated with beta-hydroxybutyrate production, showed that glucose concentrations are negatively cross-correlated with beta-hydroxybutyrate concentrations, and found that estimated net energy balance is negatively cross-correlated with beta-hydroxybutyrate. Analyses suggested that serum glucose and cholesterol concentrations significantly decreased with increased milk yield; analyses also suggested that elevated beta-hydroxybutyrate concentrations were associated with decreases in milk yield 6 d later. beta-Hydroxybutyrate and FFA concentrations 3 to 9 d before parturition were positively related to cholesterol concentrations, and relationships were positive between estimated net energy balance and cholesterol, suggesting that cholesterol concentrations increased as precursors for cholesterol synthesis were available. Time series cross-correlation analysis was a useful tool in examining relationships among variables when repeated samples were obtained from the same individuals.